Discrete events simulation to integrate operational interruption events in

block cave production scheduling

Enrique Rubio1 and Sebastián Troncoso2

Mining Engineering Department, University of Chile.

Av. Tupper #2069, Santiago, Chile.
Phone: (56-2) 978 44 99 - Fax: (56-2) 978 49 85
1
Assistant Professor, University of Chile, [email protected]
2
MEng candidate, University of Chile, [email protected]

ABSTRACT

The productivity of a Block/Panel caving operation depends upon the productivity of the mining

components. A haulage system is a critical component towards assessing the maximum productivity

of a caving mine. Traditionally, the productivity of a haulage drift depends upon the distance to the

underground crusher and the operational settings of the haulage equipment. The research

summarized in this paper shows how the secondary breakage activity carried out on the production

level could influence the productivity of a caving operation. A discrete event simulation model has

been constructed in order to quantify the productivity of a haulage drift as a result of draw point’s

hang up frequency, production crosscut’s closure strategy, and the number of secondary breakage

crews. This model has been used to simulate different scenarios in order to feed a production

nomogram that could be used to assess the true production capacity of a haulage system,

integrating operational upsets. The research has been tested in a case study in which the

productivity of a mine has been compared against estimates with and without the nomogram. The

results of including the production haulage constraint have improved the plan deviation from 51%

down to 17%.

INTRODUCTION

Mine planning is the activity that defines when and how the economic resources present in a
mineral deposit will be extracted. Its importance is vital to the development of the mineral industry
because allows linking the strategic objectives of a given mining company with the mineral
resource. One of the main tasks of the planning is to generate the production schedule which defines
the tonnages and grades to be extracted during the life of the mine, resulting finally in a
quantification of the economic potential of a mineral deposit in time. For this reason, the planning
process is betting not only for a certain productive behavior at mine level, but also jeopardizing the
economic future of the complete site.

Mining in general and particularly block cave mining can be mirrored as a piece of machinery that
take the rock mass and delivery it partitioned to posterior metallurgical benefit activities. Due to the
uncertainty in the knowledge of this mineral body, many times in the process of production
scheduling it must be considered empirical assumptions about the behave that the resource it’s
going to have inside this rock factory. Obviously, the proposed schedule is done to be reached by
operations, but generally those empirical assumptions doesn’t consider the invaluable information
collected by operations, producing that the original schedule be almost unreachable.

Particularly relevant in the commitment of a given production plan are the operational interferences,
that depending on its magnitude, the resources availability to solve them, operative strategies, etc.
impact differently in the productive behavior. To study the impact of these operational interferences
on productivity, was implemented in Arena Simulation Software, a core production unit that is
affected over time by different operational interruptions which occurrence was introduced with a
certain probability, obtained from operational data back analysis. The simulation aims to quantify
the maximum potential productive capacity of this unit, to generate a restriction on the planning
process that consider operational interferences, quantifying the maximum tonnage that a haulage
crosscut is capable to sustain in a given time-bound framework.

BACKGROUND

An underground hard rock mine constitutes a production system that consists of several
excavations. These excavations are the components of the mine production system. These
components can be the stopes, the drifts, the ore passes, etc. [1]. The productive components of a
mining system can be considered, in general, as repairable systems due to those, when a failure
occurs, can be restored to an operating condition by some repair process other than replacement of
the entire system [2]. However, it is important to find a good measure of the “age” of the
component, which is not necessarily time, in order to quantify, using the component’s productive
history, when the interruption events had occurred. To understand the impact in productivity of
interruption events, first is necessary to generate a variable that can incorporate the fact that the
mining system is imperfect in terms of productivity and it has to deal with interruption events that
occur stochastically. This variable, for this study, is the tonnage between events (TBE) and
corresponds to the tonnage that passes through a mining component between two consecutive
interruption events. TBE and its inverse (events occurrence frequency) are the most basic variables
used in this study.

SIMULATION MODEL

The simulation model used in this study is described in its most general form as a computational
discrete event stochastic simulation.

The mining infrastructure considered in the simulation model was the following:
 • Draw points (20 per production crosscut, 160 in the entire model);
• Production crosscuts (8 in the entire model);
• Ore passes( 1 per production crosscut, 8 in the entire model);
• Chutes (4 in the entire model, due to the existing “Y” connection between 2 ore passes);
• Haulage crosscuts (2 in the entire model, with 2 chutes each).

The following figure shows how the above components were connected in order to emulate the
actual mining system studied:

Figure 1: Connectivity relations considered in the simulation model.

The operational entities considered in the simulation model were the following:

• Mineral;
• LHDs;
• Dumper trucks;
• Secondary reduction crews;
• Chutes reparation crews.

Simulation model logic

The simulation model logic is based on the following rules:

• Draw point’s events occurrence is independent from other draw points in the model and
follows a given tonnage between events distribution, sampling it each time that a bucket is
drawn from a specific draw point;
• Ore passes and chutes events occurrence is independent from other ore passes and chutes in
the model, its events occurrence frequency is fixed;
• It is allowed only one LHD (modeled as an entity transfer) per production crosscut, this
equipment is specified by its velocity and its bucket size;
• The number of dumper trucks that can operate on a single haulage crosscut it is not limited,
but a specific truck must be fully loaded on a single chute;
• Only one dumper truck can load in a specific chute at the same time.
• No mechanical interruptions are considered for LHDs and dumper trucks.
 • Secondary reduction crews repairs draw points and ore passes, chutes reparation crews
repair chutes exclusively;
• Draw points in a single production crosscut are repaired by only one secondary reduction
crew each time the production crosscut is closed;
• Ore passes has a higher priority for reparations than draw points;
• Each production crosscut is specified by the distance of each draw point to the respective
ore pass;
• Each haulage crosscut is specified by the distance of each chute to the dumping station.

Simulation model validation

The following graph shows the monthly productivity evolution given by the simulation model:

Figure 2: Simulated haulage crosscut productivity evolution for validation.

In order to construct the previous graph, 10 replications of 1 year of duration each one were made in
the simulation model. The diamonds correspond to a specific replication (month to month); the
squares correspond to the monthly productivity observed in period 2005 in a specific haulage
crosscut. It can be observed that in almost every month exists al least one replication that can
reproduce the total obtained productivity. The following graph shows the scatter between the actual
productivity (of an specific haulage crosscut) and the best fit replication:

Figure 3: Best fit replication vs. Actual productivity scatter plot.

The below expression measures the differences between the Identity function and the observations,
i.e. allows quantifying the adjustment quality:

2
 
n ∧
∑  Xi − Xi 
 
s = i =1 = 10992 Kt / mes (1)
n−2

Where:

• n : Total number of points;

• Xi : Observed tonnage per month i;
∧
• Xi : Simulated tonnage per month i.

The previous value for S corresponds to an 8% of the actual average monthly productivity in period
2005. The following graph shows the scatter between the planned and actual tonnage obtained in
period 2005, for the analyzed haulage crosscut:

Figure 4: Scheduled tonnage vs. Actual productivity scatter plot.

It can be observed clearly that the schedule required productivity overestimates the actual
productivity capacities: every point in the above graph is located under the identity function
(S=51% of the actual average monthly productivity).

Nevertheless, although the best fit replication seems to be a better predictor for the monthly
productivity, it is impossible to know, before to execute the schedule, which of the replications is
the best fit, so the replication average productivity is proposed as estimation. The following graph
shows the scatter between the actual tonnage (2005) and replication average productivity (which
can be known before to execute the schedule):
Figure 5: Replication average productivity vs. Actual productivity scatter plot.

It can be observed that the adjustment of the replications average is a better indicator of the actual
productivities because the value of S (S=17% of the actual average monthly productivity) is lower
than in the case plan-actual. Nevertheless, it must be considered that the simulation considers a
fixed repair in all the period, fact that not necessarily happens in the reality, reason why the
adjustment for other haulage crosscut or other strategies could be different.

CASE STUDY

Particularly for this study were considered the following interruption events at draw points:

• Hang ups.
• Oversize.
• Wet muck.
• Compacted muck.

The following histogram shows the distribution of the values of TBE obtained from the historical
operational data analysis. It is important to note that each draw point can have many values for this
variable in its history:

Figure 6: TBE histogram.

Failure domains
The basic idea of a failure domain is to identify different sectors at the production level which are
more susceptible to interference events occurrence. So, every single draw point was characterized
by its own average TBE. The following histogram shows the distribution of the average TBE from
historical data analysis:

Figure 7: Average TBE histogram.

Naturally, every draw point has its own average TBE, but eventually, could be possible to find
components with similar tendency to the events occurrence and eventually different sectors (failure
domains) at the production level, if draw points with similar behaviors are located near enough.

One possible way to identify failure domains (and the used in this study) is to split the previous
histogram into three parts, considering the 33% and 66% quantils. The following table resumes the
average TBE limits for these three categories:

Table 1: TBE categories.

Category Lower bound (t) Upper bound (t)

Low TBE 0 1226
Medium TBE 1226 3417
High TBE 3417 52516

It can be seen in figure 8 that, although the categorization used is purely based on statistics (the
quantils no consider a space location), draw points in the same category are located, generally, in
the same zones. Considering the previous, the following failure domains are proposed:
Figure 8: Average TBE categorization (left) and failure domains (right) in the production level.

Once the failure domains are defined, in order to generate the inputs to the simulation model was
necessary to compute the distribution of TBE of each failure domain, considering every founded
value, for every draw point, for each sector. The below histograms shows the TBE distribution for
every failure domain:

Figure 9: TBE histograms for failure domains A, B and C.

Consequently, to introduce the below distributions in the simulation model, it is necessary to fit a
distribution model:

Table 2: Considered distribution models for failure domains A, B and C.

Sector Model Mean (t) Std. Dev. (t)

A Lognormal 921 1737
B Lognormal 2048 3768
C Lognormal 5320 8050

SIMULATION RESULTS ANALYSIS

Several experiments were tested in the simulation model in order to analyze the impact of the
following variables:

• Events occurrence frequency at draw points (failure domains).

• Events occurrence frequency at ore passes and chutes.
• Critical number of draw point blocked to initiate the secondary blasting activities (closure
strategy).
 • Number of secondary reduction crews.

The following graphs are representative, and help to analyze the impact of the events occurrence
frequency at draw points and ore passes and chutes, according to the simulation model results:

Figure 10: Impact on productivity of draw points events occurrence frequency (left) and impact on productivity of ore
passes and chutes events occurrence frequency (right).

The previous graph (left) shows that exist important differences in the maximum potential
productivity of the haulage crosscut due to related draw points have different events occurrence
frequency (each failure domain has its own frequency). In this aspect can be inferred that sectors
with a higher tendency to be interrupted has a lower productivity than those sectors with a lower
events occurrence frequency, what is intuitively correct understanding that a sector with a greater
events occurrence frequency passes a greater proportion of time in repairs.

Due to the infrastructure connectivity relations of the studied mining system, ore passes and its
chutes are a fundamental piece for the system’s productivity, principally because ore passes connect
that components that enter tonnage to the system (draw points) and those that extract tonnage from
the system (haulage crosscuts). There were studied 3 different events occurrence rates in ore passes
and chutes: 30Kt/event, 45Kt/event and 60Kt/event. At the same time, it must be clear that due to
the “Y” connection between 2 ore passes and a chute, to specify the same events occurrence rate in
ore passes and chutes doesn’t mean that the components fail at the same time; it means that a chute
is interrupted 2 times more often than one of the ore passes related. In figure 10 (right), it can be
seen that productivity increases while events frequency in ore passes and chutes decreases, what is
obvious due to every ton moved from draw points has to be passed through ore passes to reach the
haulage level. In the other hand, increments in productivity decrease while TBE in ore passes and
chutes increase, what is due to exist other components and entities that limit the system
productivity, and it should always be that the maximum potential productivity converge to some
value.

The following graphs are representative, and help to analyze the impact of the closure strategy and
the secondary reduction crew’s number, according to the simulation model results:
Figure 11: Impact on productivity of production crosscuts closure strategy (left) and impact on productivity of secondary
reduction crew’s number (right).

It can be seen in figure 11 (left) that the production crosscuts closure strategy (critical number of
draw points detained for the production crosscut closure for draw points clearance) impacts
differently depending on the secondary reduction crew’s number, in the analyzed time scale
(monthly productivity). When the crew’s number is reduced there are no big differences in the
system productivity by using a different closure strategy, and vice versa; when the total number of
secondary reduction crews is higher, the closure strategy becomes more important in terms of
productivity.

It is important to note that secondary reduction crews is a common resource in the entire mine, issue
that the simulation model is incapable to reproduce due to only emulates a basic productive unit and
not the complete mining system. However, it is possible to observe in the obtained results the
impact of the total available resources (secondary reduction crews) in productivity, principally
because the simulation model is bigger enough (128 draw points distributed in 8 production
crosscuts with one ore pass each), so the secondary reduction crews are not, in general, a redundant
resource in the simulation model. Nevertheless, before to extrapolate the obtained results is
necessary to normalize them, i.e. to understand that when the model says 4 secondary reduction
crews, these was assigned to a total of 8 production crosscuts, so is more precise to say 0.5
secondary reduction crews per production crosscut, to extrapolate the results to the entire mine.

In figure 11 (right) it can be seen that the haulage crosscut productivity is direct related with the
secondary reduction crew’s number. The greater is the secondary reduction crews number, the
bigger the haulage crosscut productivity is, which is intuitive understanding that secondary
reduction crews returns to production those infrastructure components which are down due to
operative interferences.

In turn, it can be seen that increases in productivity, as more secondary reduction crews are
dwindling, reaching a limit above which no productivity changes with the increase of secondary
reduction crews are observed. This is mainly due to the productive capacity of the system is not
limited only by operational interferences, which are handled by the crews, but also by other factors
such as the productivity of LHDs located upstream, so naturally there is a productivity limit on
which secondary reduction crews begin to be underutilized because they are redundant.
PRODUCTION NOMOGRAM

As a result of the simulation model integrating defined failure rates, a production nomogram is
proposed to constraint the maximum haulage tonnage through put. Certainly this production
constraint would influence the production target as a result of a mine planning cycle.

The proposed nomogram allows relating the maximum potential productivity of a haulage crosscut
as a function of the events rate at draw points (failure domains), the closure strategy at production
crosscuts for draw point’s clearance, the number of secondary reduction crews and the events rate at
ore passes and chutes.

Figure 12: Productivity nomogram.

The following table related the secondary reduction crew’s number and the operational strategy
(critical draw points) with the entrance parameter of the nomogram (normal productivity) and the
adjustment factor, for each sector:

Table 3: inputs parameters for productivity nomogram.

Red. crews ADJUSTMENT FACTOR

number / Normal productivity
SECTOR TBE Ore passes and chutes
Critical draw (Events/month)
points 30 (Kt/month) 45 (Kt/month) 60 (Kt/month)
1/1 A/B/C 17.5 / 22.86 / 37.54 0.1 / 0.1 / 0.1 0.104 / 0.106 / 0.104 0.106 / 0.109 / 0.109
1/4 A/B/C 19.42 / 28.11 / 50.77 0.1 / 0.1 / 0.1 0.102 / 0.106 / 0.107 0.108 / 0.11 / 0.11
1/7 A/B/C 20.06 / 29 / 54.63 0.1 / 0.1 / 0.1 0.105 / 0.108 / 0.106 0.107 / 0.112 / 0.114
1 / 10 A/B/C 20.38 / 29.94 / 56.97 0.1 / 0.1 / 0.1 0.103 / 0.106 / 0.106 0.106 / 0.11 / 0.113
1 / 13 A/B/C 20.16 / 29.95 / 57.94 0.1 / 0.1 / 0.1 0.105 / 0.107 / 0.108 0.108 / 0.111 / 0.114
3/1 A/B/C 45.74 / 52.64 / 93.23 0.1 / 0.1 / 0.1 0.106 / 0.116 / 0.108 0.108 / 0.121 / 0.113
3/4 A/B/C 50.09 / 68.37 / 93.58 0.1 / 0.1 / 0.1 0.106 / 0.109 / 0.135 0.109 / 0.113 / 0.15
3/7 A/B/C 51.5 / 72.78 / 96.94 0.1 / 0.1 / 0.1 0.105 / 0.107 / 0.137 0.109 / 0.111 / 0.152
3 / 10 A/B/C 52.67 / 76.18 / 101.15 0.1 / 0.1 / 0.1 0.103 / 0.108 / 0.133 0.107 / 0.112 / 0.149
3 / 13 A/B/C 52.47 / 80.24 / 102.08 0.1 / 0.1 / 0.1 0.104 / 0.107 / 0.133 0.108 / 0.111 / 0.15
5/1 A/B/C 49.47 / 73.11 / 92.19 0.1 / 0.1 / 0.1 0.11 / 0.107 / 0.135 0.123 / 0.116 / 0.158
5/4 A/B/C 61.36 / 92.48 / 92.15 0.1 / 0.1 / 0.1 0.121 / 0.116 / 0.138 0.126 / 0.118 / 0.171
 5/7 A/B/C 69.1 / 95.26 / 94.34 0.1 / 0.1 / 0.1 0.114 / 0.12 / 0.138 0.118 / 0.126 / 0.172
5 / 10 A/B/C 75.32 / 98.67 / 101.49 0.1 / 0.1 / 0.1 0.108 / 0.118 / 0.131 0.111 / 0.123 / 0.161
5 / 13 A/B/C 81.25 / 101.55 / 100.88 0.1 / 0.1 / 0.1 0.103 / 0.116 / 0.135 0.106 / 0.121 / 0.163

Finally, the maximum potential productivity of the haulage crosscut is obtained as follows:

MPP( Kt / month ) = IP(events / month) * TBEOREPASSES _ CHUTES ( Kt / event ) * AF

(2)

Where:

• MPP: Maximum potential productivity of the haulage crosscut;

• IP: Intrinsic productivity;
• TBEOREPASSES _ CHUTES : Tonnage between events in ore passes and chutes;
• AF: adjustment factor.

For example, for sector A, with 3 secondary reduction crews and 13 critical draw points, from the
previous table it is read a normal productivity equal to 52.47events/month and an adjustment factor
of 0.108, if it is consider a TBE in ore passes and chutes of 60Kt/event, from the previous
nomogram it is read an intrinsic productivity of 26.3event/month, so we get the following MPP
value:

MPP( Kt / month) = 26.3(events / month) * 60( Kt / event ) * 0.108 = 170.4( Kt / month)

(3)

CONCLUSIONS

The most relevant variables in the haulage crosscut productivity are at least the following:

• Interruption events occurrence frequency (at different infrastructure components at the

production and haulage levels);
• Number of secondary reduction crews;
• Closure strategy of production crosscuts for its reparation.

In relation to the previous points, from the results of the simulations experiments were found that:

• Haulage crosscuts productivity is inverse related with events occurrence frequency in draw
points, ore passes and chutes. Particularly for the case study were observed differences
from 23% to 67% in the monthly productivity of haulage crosscuts for different events
occurrence frequencies at level of draw points and from 4% to 40% for different events
occurrence frequencies in ore passes and chutes;
• A higher number of secondary reduction crews are related with a higher productivity, with
an upper limit that depends at least on the events occurrence frequency in the different
mining components. Monthly productivity in haulage crosscuts varies from 3% to 77% with
changes in this variable;
 • The closure strategy that maximizes the haulage crosscut productivity depends at least on
the events occurrence frequency and the number of secondary reduction crews. Monthly
productivity in haulage crosscuts varies from 1% to 35% with changes in this variable.

Related to the previous points, the following issues are of interest for the production scheduling
process:

• It must be understand in the planning process that the mining infrastructure has a limited
production capacity and this limit not necessarily depends only on the draw point’s
productivity;
• It must be considered operational standards for the secondary reduction (and others
activities for infrastructure components reparation) according with the production
requirements.

ACKNOWLEDGEMENTS

To División Andina, Coldelco Chile to contribute with their production data and expertise to
implement the simulation model.

To the University of Chile to support the publication of the results associated to the research
presented in this paper.

REFERENCES
Kazakidis, V. N. and Scoble, M. (2002) Accounting for ground-related problems in mine production systems planning,
Mineral Resources Engineering. Vol 11, No.1, Imperial College Press, pp. 35-57. [1]
Rigdon, S. and Basu, A. (2000) Statistical Methods for the Reliability of Repairable Systems, John Wiley & Sons,
Toronto, Canada, pp. 1-22. [2]
Troncoso, S. (2006) Simulación del Impacto de Interferencias Operacionales para la Planificación de Producción,
Universidad de Chile, Santiago, Chile. [3]