Mining Technology

Transactions of the Institutions of Mining and Metallurgy: Section A

ISSN: 1474-9009 (Print) 1743-2863 (Online) Journal homepage: http://www.tandfonline.com/loi/ymnt20

Application of discrete event simulation in

optimising coal mine room-and-pillar panel width:
a case study

Angelina Anani, Kwame Awuah-Offei & Joseph Hirschi

To cite this article: Angelina Anani, Kwame Awuah-Offei & Joseph Hirschi (2016): Application
of discrete event simulation in optimising coal mine room-and-pillar panel width: a case study,
Mining Technology, DOI: 10.1080/14749009.2016.1195035

To link to this article: http://dx.doi.org/10.1080/14749009.2016.1195035

Published online: 20 Jun 2016.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=ymnt20

Download by: [Missouri S & T] Date: 21 June 2016, At: 12:33

 Application of discrete event simulation in
optimising coal mine room-and-pillar panel
width: a case study
Angelina Anani1*, Kwame Awuah-Offei2 and Joseph Hirschi3
A key design aspect of room-and-pillar coal mines is the panel width (or number of entries in a panel),
which affects unit mining costs and productivity. Traditional mine design approaches do not facilitate
optimisation of unit mining costs and productivity as a function of the panel width. Discrete event
simulation can be used to facilitate optimal panel width selection that minimises unit mining costs and
maximises productivity. The objective of this study is to evaluate the impact of panel width on the cost
and productivity of a room-and-pillar operation using discrete event simulation. The mining system is
modelled as a discrete event model that estimates unit costs and productivity for a given panel width
Downloaded by [Missouri S & T] at 12:33 21 June 2016

(number of entries). Data from a real-life mine is used to validate the model in Arena®. An optimal panel
width is recommended based on simulation results.
Keywords:  Discrete event simulation, Room-and-pillar mining, Panel width optimisation, Simulation analysis program, Monte Carlo simulation, Fleet
size optimisation, Coal mine, Productivity

Introduction The traditional approach for determining panel width focuses

Room-and-pillar (R&P) mining is one of the oldest self-­
supported underground mining methods used for exploiting
flat and tabular deposits such as coal. One of the main goals of
R&P mine design is to extract the maximum possible amount
of ore while maintaining the bearing rock’s strength and con-
dition. Design parameters in R&P mining depend on several
factors, including depth of mining, production recovery, sta-
bility of the hanging wall and coal strength (Farmer 1992). A
key aspect of R&P mine design is panel design. Panel design
relies on pillar strength and dimensions within the panel, as
well as coal recovery and production requirements.
Paneling in R&P mining is done to divide the mine into
different areas for ventilation and other aspects of mine oper-
ation. Barrier pillars are used to separate panels thus pre-
venting progressive collapse in case of pillar failure within a
panel. The number of barrier pillars needed to separate panels
depends on the number of panels being planned. As panel
width increases, so does barrier pillar size. Panel dimensions
also affect the mining (cut) sequence with wider panels result-
ing in more complicated cut sequences and more tramming by
continuous miners (CMs), a necessary function during which
coal is not produced. Consequently, panel width (number of
entries) affects the overall cost and production recovery of an
operation. An efficient R&P mine design takes into account
the optimum panel width that maximises production and min-
imises cost while ensuring bearing rock stability.
1
Mining and Nuclear Engineering, Missouri University of Science and
Technology, Rolla, MO 65401, USA
2
Mining and Nuclear Engineering, Missouri University of Science and
Technology, Rolla, MO, USA
3
Mining and Mineral Resources Engineering, Southern Illinois University,
Carbondale, IL, USA
*Corresponding author, email akakc2@mst.edu
primarily on ventilation requirements, haulage distances (fleet
requirements), geotechnical conditions and property boundaries
(Dunrud 1998; Zipf 2001; Loui and Sheorey 2002, Ghasemi
and Shahriar, 2012). However, it is important that panel width
determination takes into account minimising operating costs,
optimising cut sequences, and maximising overall productivity.
Panel dimensions are determined by the number of pil-
lars and rooms in the panel. Panel width is defined by the
number of entries (production and development) excavated
in the panel. Due to advances in geotechnical methods over
the past few decades, varying designs with increasing panel
widths have been implemented (Robson 1984; Cain 1999).
Common strategies employed in R&P panel development
include: (i) developing a base width (typically seven or more
entries) on advance in increments (measured in crosscuts) that
correspond with required belt and power moves, (ii) mining
rooms on either side or both sides of the base width and (iii)
recovering pillars on retreat, if that practice is allowed.
Typically, the importance of panel design is emphasised
in retreat mining where entries and rooms are mined first and
then panel pillars are mined afterwards. Pillar recovery is
common in metal mines where high grade ore cannot be left
behind. Even though panel recovery in coal mines is not prev-
alent, recent advances in electric haulage units have spurred
a move towards wider panels to take full advantage of their
capabilities. However, the effect of wider panels on produc-
tivity and unit operating costs has not been fully investigated.
Traditional design approaches do not facilitate optimisation
of unit mining costs and productivity as a function of panel
width (Zipf 2001; Loui and Sheorey 2002; Ghasemi and
Shahriar 2012). The objective of this study was to evaluate
the impact of panel width (number of entries) on the cost and
productivity of R&P operations. The authors present R&P

© 2016 Institute of Materials, Minerals and Mining and The AusIMM

Published by Taylor & Francis on behalf of the Institute and The AusIMM
Received 24 March 2015; accepted 24 May 2016
DOI 10.1080/14749009.2016.1195035 Mining Technology   2016 1
 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width
panel width design as a possible dual objective optimisation
problem where we seek to maximise productivity and mini-
mise unit costs subject to all constraints. The approach is to
build a discrete event simulation (DES) model and use it to
conduct experiments to estimate the productivity and unit
costs at varying panel widths and operating conditions. As a
case study, a coal mine is modelled as a DES using Arena®
(Rockwell Automation Inc., Milwaukee, WI).
Discrete event simulation for panel
width optimisation
DES for engineering design
DES is a computer-based approach that facilitates modelling,
simulation and analysis of the behaviour of complex systems
as a sequence of discrete events. DES software has been con-
tinuously improved for the past four decades leading to more
advanced simulation languages (Nance 1995; Pegden et al.
1995; Rice et al. 2005). Simulation languages are symbols/
codes recognised by computers or computer programs as
issued commands a programmer wishes to perform (Kiviat
1969). Common simulation languages currently used for DES
include SIMAN (system modelling), GPSS and SLAM. This
work uses Arena®, which is based on the SIMAN language
Downloaded by [Missouri S & T] at 12:33 21 June 2016
for DES modelling and simulation.
SIMAN is a SIMulation ANalysis program generally used to
model either discrete, continuous or a combination of discrete
and continuous systems (Pegden et al. 1995). SIMAN allows
process-oriented, event-oriented and continuous components
to be integrated into a single system. A unique characteristic
of a SIMAN program is the distinct decomposition of model
and experimental frames. The static or dynamic nature of a sys-
tem can be defined in the system model. Different experiments
can be done in the experimental framework resulting in mul-
tiple sets of output (McHaney 1991). However, the close link
between its arithmetic and list processes on the one hand and
its demand-resource concepts on the other restrict its capability
to model demand-driven systems (Fishman 2001).
Applications of discrete event simulation as a
­decision-making tool for improve mining systems are vast
and increasing (Doe and Griffin 1987; Sturgul and Harrison
1987; Harrison and Sturgul 1988; Basu and Baafi 1999;
Vagenas 1999; Awuah-Offei et al. 2003; Michalakopoulos
et al. 2005; Yuriy and Vayenas 2008; Ben-Awuah et al.
2010). However, DES application in R&P mining is limited
to a few examples (Hanson and Selim 1974; Pereira et al.
2010). None of them deal with using DES to determine
optimal design parameters. Hanson and Selim (1974) used
event based models to compare room and pillar mining
with long wall mining systems. Pereira et al. (2010) used
discrete event simulation to maximise the coal faces extrac-
tion from a given panel. It still remains that DES applica-
tion specifically in room and pillar mines is limited even
more so in optimising design parameters.
DES can be used as a decision-making tool in determining
optimum design parameters. Since DES can be used to simu-
late system performance at varying operating conditions and
design parameters, what-if analysis can be performed quickly
and cheaply with a valid model. Through such experiments,
optimum design parameters can be determined that meet design
goals and respect all constraints of the design problem. For
instance, to design a greenhouse crop system for maximum
production and quality of labour, Van’t Ooster et al. (2013) suc-
cessfully used DES to perform sensitivity analysis in identify-
ing parameters that influence labour performance and the effect
2 Mining Technology   2016
of uncertainty on the performance indicator. Similarly, Petering
(2009) investigated the optimal width of storage blocks in a
terminal container and its effect on gross crane rate.
The advantage of DES lies in its ability to model complex
systems with relative ease. DES allows for implementation of
new designs and methods without interfering with the real-life
system. DES also helps answer the question of why certain
phenomenon occur (Asplund and Jakobsson 2011). It has the
ability to capture random behaviour caused by a large number
of factors that impact the system, using the Monte-Carlo sim-
ulation technique. This gives a sense of interactions between
variables that make up such systems.
DES can be used to perform ‘bottleneck’ evaluations to
­discover where work in process in a system is delayed and
which variables affect this. Identifying problems and gaining
understanding into the importance of these variables increases
awareness of their importance relative to the performance of the
overall system. DES allows an analyst to vary the system oper-
ating periods, cheaply and easily (Schriber 1977). On the other
hand, even though DES provides a way to analyse and under-
stand the changing behaviour of the system, it only provides an
estimate of the model output. Building DES models can be costly
and time consuming. Therefore, it is used on an as-needed basis
where benefits outweigh costs (Asplund and Jakobsson 2011).
This study seeks to apply DES to facilitate better panel width
design. The general concept is presented in Section DES for
panel width design
DES for panel width design
Ultimately, optimising a design parameter is an optimisation
problem as described by Equation (1). The decision variable,
vector x, represents variables that affect the objective func-
tion, f (x). Possible values that these variables can take make
up the set of feasible solutions (alternative designs).
minimise f (x)
(1)
subject to x ∈ Ω
In the case of panel width design, the objective function could
reflect the desire to maximise mining recovery and produc-
tivity as well as minimise unit operating costs. This work
focuses on the dual objective of maximising productivity and
minimising unit costs. Decision variables can be panel width,
cut sequences, the number of CMs and the number of haulage
units assigned to each CM.
If the objective function can be written mathematically
(explicit) in terms of the decision variables and all constraints can
be described similarly, there are many techniques to solve such
optimisation problems. Often, however, the objective function is
highly non-linear and implicit. In such cases, very few techniques
(simulation being one) can solve the problem (Kleijnen 1998).
In the case of panel width optimisation, productivity and
unit costs associated with production functions of cutting,
loading and hauling as a function of the panel width, equip-
ment fleet and cut sequence are non-linear and implicit. DES
offers a means to estimate the unit cost and productivity for
a given panel width, equipment fleet and cut sequence. The
approach taken in this work is to first build a valid DES model
of coal loading and hauling operations, then determine a feasi-
ble set of decision variable values (panel widths, fleet and cut
sequences) and estimate all productivity values and unit costs
for these conditions using the model. The optimal solution
can then be selected on the basis of the objective function.
To find the optimal solution, one would have to determine
the relative significance of productivity and unit cost to the
 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width
1 Haulage unit dumping time
Downloaded by [Missouri S & T] at 12:33 21 June 2016
2 Empty haulage unit travel speed
3 Loaded haulage unit travel speed
4 Loaded haulage unit travel time
decision. Since this varies from one situation to another, the
authors chose not to attempt finding a single optimal solution,
but to present a discussion of results relative to productivity
and unit costs.
 Mining Technology  2016
Case study
A case study of an actual coal mine is presented in this section
to illustrate the approach discussed in Section DES for panel
width design. The discussion here follows the general steps
of a typical simulation study (Kelton et al. 2003).
Problem formulation
The objective of the panel width optimisation study is to eval-
uate the impact of panel width on the unit cost and produc-
tivity of an underground R&P operation. A DES model with
variables that characterise the mining system was built using
Arena®. The model predicts unit mining costs and produc-
tivity at different panel widths. The DES model was validated
with shift production data obtained from a R&P coal mine
in southern Illinois, United States of America (USA). The
defined performance metric was that the relevant simulated
output should be within 15% of actual values from the mine.
System and simulation specifications
The mine used for this study is located in southern Illinois,
USA. The mine produces approximately 7  million tons of
coal per year from the Herrin No. 6 seam using R&P mining
methods with a panel recovery rate of 54%. Eight Joy Model
14CM27 CMs (two for each panel) cut and load coal at up to
40 tons per minute with a maximum cutting height of 11.2 feet.
Coal is hauled from CMs to feeder-breakers by 20-ton Joy
Model BH20 battery-powered haulage units. A ­feeder-breaker
is located at the centre of each production panel to transfer
mined coal from haulage units to conveyor belts. As the panel
advances, it is moved forward in three-crosscut increments
with each crosscut forming a row of in-line pillars across the
panel. The full width of the panel is mined in six-cross-cut
increments. The mine has experimented with different panel
widths and mining sequences. Currently, advancing base
widths of 11 and 13 entries before mining rooms is the most
common. Maximum and minimum panel widths are 21 and
11 entries, respectively. Each CM mines up to seven entries
on one side of a panel.
The objective of the simulation is to develop a valid DES
model that predicts unit mining costs and productivity and
provides basic animation for verification. Input data used in
the model were obtained from time studies done at the mine as
described by histograms in Figs. 1–6. Raw data were analysed
to fit statistical distributions using the chi-squared goodness-
of-fit test as shown in Table 1. Input data include loading and
dumping times, payloads and battery change data, which are
sampled from the distribution. Model output includes produc-
tion per shift, tons per hour, total operating costs including
equipment costs and the calculated cost per ton for a given
panel width.
Model formulation: CM and haulage logic
The DES modelling framework requires system entities,
resources and processes to be specified by the analyst. To
initiate modelling, entities go through defined processes in
a logical manner waiting for needed resources to become
available at each process (i.e. resources are ‘busy’ if they are
being used by other entities) before they go through the pro-
cess. The CM is modelled as a resource used for the loading
process and can only load one haulage unit at a time. Loads
of coals are modelled as entities with specific attributes (entity
3
 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width

number, payload and cut sequence). Battery-powered haulage

units are model as guided transporters used for hauling loads
(entities). A guided transporter is an Arena®-specific mod-
elling construct for material haulage (Rockwell Automation
Inc. 2012). Transporters use entries and crosscuts as haulage
routes, which are modelled to restrict traffic flow such that any
point on a haulage route can only accommodate one haulage
unit at a time since mine openings are not wide enough for
them to pass each other. The feeder-breaker is also modelled
as a stationary resource used for dumping loads (entities). The
feeder-breaker and each cutting face are modelled as stations.
Points in the model where transporters transfer entities are
called stations. Haulage routes between stations are mod-
5 Haulage unit spotting time elled as network links to capture varying haulage distances.
Distances for each network link are an input to the model.
Figure 7 shows the logic used to model the system.

Verification and validation

An animation of the system was designed and used to verify
that the model performs as intended. The resource, trans-
porters, stations, and network links are modelled as part of
the animation for loading and transporting coal (entities).
Shift production data from the mine was used to validate
Downloaded by [Missouri S & T] at 12:33 21 June 2016

the model. For validation, the simulation model predicted

coal production (load count/shift) and shift duration, which
was compared with data from a time-and-motion study con-
6 CM travel time between cuts ducted in one of the sections of the host mine where the panel
was being advanced with 13 entries. The time-and-motion
study collected data for 18 CM cuts completed over two

Table 1 Input data

Data (s) Distribution (s) P-value

Payload (kg) 10,886 –
Empty speed (m s−1) 1.86 + GAMM(0.0987, 5.05) <0.005
Loading time (s) 28 + ERLA(3.63, 3) <0.005
Dumping time (s) 6 + GAMM(2.79, 5.36) <0.005
Battery change (s) TRIA(5, 7, 10) <0.005
Loaded speed (m s−1) 1.77 + GAMM(0.0546, 7.26) <0.005
Time between cuts (s) NORM(797, 87.7) <0.005
Spotting time (s) 12.5 + GAMM(4.22, 2.11) <0.005

7 DES model logic

4 Mining Technology   2016

 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width
8 Cut sequence for 11-entry base width
non-consecutive shifts. During the first shift, 6.33 hours were
spent making 11 cuts with the remaining time spent on con-
veyor belt and CM repairs. The second shift was spent entirely
Downloaded by [Missouri S & T] at 12:33 21 June 2016
on production; however, data was collected for only the first
half of the shift during which seven cuts were completed.
For both shifts, coal was hauled by four haulage units with
an average payload of 10,886 kg (12 tons). The production
reported from the CM’s onboard monitoring system indicates
that the mine produced 2.2M kg (2448 tons) of coal from 204
loads and 3.2M kg (3576) from 298 loads during the first shift
and second shift, respectively.
In the validation experiment, 150 replications were con-
ducted to obtain estimates of load count and total coal pro-
duction, mining duration and other output. The number of
replications selected was such that the half-width1 of the min-
ing duration (the most uncertain output) is less than 1% of the
estimated duration. The cut sequence used in the validation
experiment duplicated that used during the time-and-motion
study. Each replication stops when all specified cuts have been
mined in the simulation. Results are discussed in the Results
and Discussion section.
Simulation experiments and analysis
Simulation experiments were designed to analyse the effect
of panel width (number of entries), number of haulage units
assigned to each CM, and the cut sequence on mining cost
and productivity. At the mine, the staff have experimented
with cut sequences that advance 11 or 13 entries first before
expanding into rooms, if necessary. Hence, these experiments
were to evaluate whether to advance with 11 or 13 entries
before mining rooms. Once the initial advance is mined,
the mine has mined anywhere from no additional rooms to
five additional rooms on each side. Hence, the experiment
includes three factors:
• Number of base width entries (11 or 13);
• Number of rooms (0, 1, …, 5 rooms on each side of the
panel); and
• Number of haulage units assigned to each CM (3, 4,
or 5).
This leads to a total of 2 × 6 × 3 = 36 combinations of
experiments. For each experiment, 150 replications were run
for the analysis. Each replication was run until all cuts in the
sequence have been mined. The cut sequence was provided
 Mining Technology  2016
as an input based on mining practices at the mine. Mining
faces in the 11- or 13-entry base width are mined using the
cut sequence shown in Fig. 8. Rooms are mined using the
cut sequence shown in Fig. 9. The experiment evaluates a
super-section mining system with two CMs (one on each side
of the section). The conveyor belt is located in the centre
entry of the panel.
The simulation output includes production data (e.g. load
count and total production tonnage), duration of mining and
percentage of time the CM spends loading haulers. Other
outputs include total cost of mining and estimated unit costs
(Equation (2)). Results of simulation experiments are dis-
cussed in the Results and Discussion section.
nCM CCM + nH CH tr + CF
( )
Unit costs($∕ton) = 2
Total production
where nCM and nH are the number of CMs and haulage units,
respectively; tr is the duration of the simulation run; CCM and
CH are hourly costs for CMs and haulage units, respectively;
and CF is fixed costs, which include labour and equipment for
advancing belt and power systems.
Results and discussion
Validation
Table 2 shows the results of the validation experiments for the
first production shift. The model takes a bit longer (30 min
more) to mine the 11 cuts and also loads 24 more haulage units
than the observed system. The key performance measures are
the number of loads mined from the 11 cuts and the duration,
which are within 11 and 8%, respectively, of the actual values.
Both values are within the 15% specified earlier. The model
was thus deemed valid and used for all the experiments.
Effect of panel width
Figures 10–17 show simulation results for experiments with
the default number of haulage units (four per CM) where 11-
and 13-entry systems are mined with 0, 1, 2, 3, 4 or 5 rooms
on each side for a total number of entries between 11 (11-entry
system with 0 rooms on each side) and 23 (13-entry system
with 5 rooms on each side). These results indicate the effect of
panel width (number of entries) on productivity and unit costs.
5
 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width

9 Room cut sequence for 11 entry advance with two additional rooms on each side

Table 2 Results of validation experiment

Parameter Actual Simulated Difference

Duration of mining (hours) 6.33 6.83 8%
Production (tons) 2448 2748 12%
Number of haulage units loads 204 226 11%
Half-width of duration (hours) – 0.012 –
Downloaded by [Missouri S & T] at 12:33 21 June 2016

11 entry system 13 entry system

Figures 10 and 11 show that total production and duration
100,000 of mining increase with increasing number of entries. This
Producton (tons)

is what one would expect, if the model is performing well.

Figures 12 and 13 show that the percentage of production
50,000 time the CM spends loading haulage units initially increases
with increasing panel width until an optimal panel width. This
0
indicates that there is excess haulage unit capacity in the sys-
tem with less than optimal number of entries. CM operations
11 13 15 17 19 21 23
are inefficient due to the excessive spotting time resulting
Number of entries in long wait times and bunching; however, expanding panel
width beyond the optimal results in inadequate haulage unit
10 Total production
capacity and under utilisation of the CM. This is confirmed
by Figs. 14 and 15 showing that the optimal panel width.
Initial expansion of the panel reduces the haulage unit cycle
11 entry system 13 entry system
time (minimises waiting time). However, further expansion
240 of the panel increases haulage unit cycle times because haul
distances become longer, leading to a haulage unit constrained
Duration (hrs)

operation. Adding more haulage units will increase produc-

140 tivity and CM utilisation as discussed in Section Effect of
number of haulage units.
These trends (cycle time and CM loading times)
40 directly result in the observed trend in productivity (Fig. 16
11 13 15 17 19 21 23 Productivity). Panel widths of 17 and 19 entries result in max-
Number of entries imum productivity when advancing with a base width of 11
entries and 13 entries, respectively. However, this trend is not
11 Duration of mining
mirrored in unit cost results (Fig. 17) due to the effect of fixed
costs that make larger panels more cost-effective even with
sub-optimal productivity. In Fig. 17, unit costs are estimated
11 entry system 13 entry system
using Equation (2). Hourly costs of haulage units and CM
25.5 are estimated at $79.13 and $122.40 (InfoMine 2013).2 Fixed
Loading time (%)

costs for moving the belt are estimated at $81,050.

25.0 The following observations can be made from these
results:
24.5 • 11-entry systems outperform 13-entry systems under
similar conditions (cut sequences and equipment);
24.0 • Haulage unit cycle times correlate very well with pro-
11 13 15 17 19 21 23 ductivity and CM loading;
Number of entries • There appears to be an optimal panel width for a given
number of haulage units based on productivity analysis;
12 CM time spent loading (LHS) and

6 Mining Technology   2016

 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width

11 entry system 13 entry system 11 entry system 13 entry system

3.40
CM Utilization (%)

Cost per ton ($/t)

25.0
2.40
24.5

1.40
24.0 11 13 15 17 19 21 23
11 13 15 17 19 21 23
Number of entries
Number of entries
17 Unit costs
13 CM time spent loading (RHS)

11 entry system 13 entry system

3 Cars 4 Cars 5 Cars
10.55
10.50 560
Cyle time (mins)

10.45

Productivity (tph)
10.40 550
10.35
10.30 540
10.25
10.20 530
11 13 15 17 19 21 23
Downloaded by [Missouri S & T] at 12:33 21 June 2016

Number of entries 520

11 13 15 17 19 21
14 Average cycle times (LHS) Number of entries

18 Effect of number of haulage units on productivity for

11 entry system 13 entry system 11-entry system
10.55
10.50
Cyle time (mins)

however, the increase when the number of haulage units

10.45
increases from three to four is much more significant than
10.40
the increase when the number of haulage units increases from
10.35
four to five. Also, the number of haulage units can affect
10.30
optimal panel width. For example, Fig. 18 shows that optimal
10.25
panel width with three haulage units assigned to each CM is
10.20
11 13 15 17 19 21 23 13 entries, whereas with four haulage units, optimal panel
width is 17 entries. This is because the number of assigned
Number of entries
haulage units affects the width at which the system becomes
limited by haulage unit capacity.
15 Average cycle times (RHS)
Figures 20 and 21 show the sensitivity of unit cost results
to number of haulage units. With each additional haulage
11 entry system 13 entry system unit, unit costs increase for both 11- and 13-entry systems.
The following observations can be made based on these
560
results:
Productivity (tph)

555 • Results are sensitive to the number of haulage units as

550 follows:
• Productivity increases with additional haulage units, and
545 • Optimal number of entries changes with varying number
540 of haulage units.
11 13 15 17 19 21 23 • Cost increases outpace productivity increases with each
Number of entries additional haulage unit leading to higher unit costs.

16 Productivity
• Unit costs decrease with increasing number of entries
due to the effect of fixed costs.
Effect of number of haulage units
Figures 18 and 19 show the sensitivity of productivity results
to the number of haulage units. It can be observed that with
the addition of each haulage unit, productivity increases;
Effect of fixed costs
From Equation (2), if fixed costs are negligible, the unit costs
curve should be the inverse of the productivity relationship.
However, Figs. 16 and 17 do not show this relationship
indicating that fixed costs significantly affect the unit cost
relationship. Figure 22 shows the sensitivity of the unit cost
relationship to fixed costs using results for the 11-entry system
with four haulage units (same as Fig. 17). Figure 22 shows
that the unit cost relationship will indeed show an optimal at
17 entries if fixed costs are less than or equal to $1000. Fixed

 Mining Technology  2016 7
 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width

3 Cars 4 Cars 5 Cars costs as low as $2000 more than compensate for any decline in
productivity due to under-resourced CMs. That is, with high
560 fixed costs (≥$2000), unit costs for mining larger panels will
be lower, even though productivity will be sub-optimal after
Productivity (tph)

550 the panel width exceeds the optimal panel width for produc-
tivity. From a cost perspective, larger panels are advantageous
540 because of fixed costs included in moving belt and power.

530
Conclusion
520 This research effort has successfully built a discrete event
13 15 17 19 21 23 simulation model that can be used to facilitate panel
Number of entries width design. The DES model is capable of ­evaluating
the effect of panel width (number of entries) on R&P
19 Effect of number of haulage units on productivity for mine ­productivity and unit costs. The DES model has
13-entry system
­successfully been ­validated for the mine that cooperated
with this study. The validated model has been used to
­evaluate the effect of panel width on productivity and unit
3 Cars 4 Cars 5 Cars
costs of the mine.
$3.60 Based on results of this work, the following general con-
clusions can be made:
Unit costs ($/ton)

$3.20 • For particular operating conditions (equipment, cut

sequence, etc.), there exist an optimal panel width that
$2.80
maximises productivity.
Downloaded by [Missouri S & T] at 12:33 21 June 2016

$2.40 • For particular operating conditions, an optimal panel

width exists that minimises unit costs, only if the fixed
$2.00 costs are negligible. For any significant fixed cost, larger
11 13 15 17 19 21 panels will always result in lower unit costs.
Number of entries For the cooperating mine, in particular, the following con-
clusions can be drawn:
20 Effect of number of haulage units on unit costs for • The 11-entry system is better than the 13-entry system.
11-entry system This is a function of cut sequences used.
• The practice of moving the belt after mining three cross-
cuts to ensure haul distances to rooms is reasonable.
3 Cars 4 Cars 5 Cars
• The optimal panel width under simulated conditions is
$3.60 17 entries (3 rooms on each side of the 11-entry base
width).
Unit costs ($/ton)

$3.20 • Four (4) haulage units should be assigned to each CM

in the panel.
$2.80

$2.40
Acknowledgement
$2.00
13 15 17 19 21 23 This work was made possible with funding from the Illinois
Clean Coal Institute. The authors are grateful for the support
Number of entries
of Prairie State Generating Company, owners of the cooper-
21 Effect of number of haulage units on unit costs for
ating mine for their support during this research. The authors
13-entry system are also thankful to Ms Sisi Que and Mr Mark Boateng for
their assistance.

$- $500.00 Funding
$1,000.00 $2,000.00 This work was supported by Illinois clean coal institute
$1.65 [13/3B-1].
Unit cost ($/ton)

$1.60

$1.55 Notes
  1. Half-width = tn−1,1−𝛼∕2 √s n, tn−1, 1−α/2 = critical values from
$1.50
11 13 15 17 19 21 t  tables, n  =  number of replications, s  =  sample
No. of entries standard deviation.
  2. InfoMine cost data is used in this paper to protect the
22 Effect of fixed costs on unit cost relationships mine’s confidential data.

8 Mining Technology   2016

 Anani et al.  Application of discrete event simulation in optimising coal mine room-and-pillar panel width
Notes on contributors
Angelina Anani is a postdoctoral fellow at the Missouri
University of Science and Technology. She holds a PhD
and BSc degree in mining engineering from the Missouri
University of Science and Technology. Her current research
interest includes modelling and optimisation of mining
systems.
Kwame Awuah-Offei is an associate professor of mining engi-
neering at Missouri University of Science and Technology.
His research interests revolve around applying modelling,
simulation and optimisation to promote sustainable mining
practices. He holds a PhD and BSc degrees in mining engi-
neering. He is a registered professional engineer.
Joseph Hirschi is an assistant professor of mining engineering
at Southern Illinois University Carbondale. His research inter-
ests are in mine and power plant optimisation; mining meth-
ods; coal processing; and miner health, safety, and training.
He holds a PhD, MSc, and BSc degrees in mining engineering
and an MBA. He has been certified by the State of Illinois
as a First Class Miner, Mine Examiner, Mine Manager, and
Shot Firer.
References
Downloaded by [Missouri S & T] at 12:33 21 June 2016
Asplund, S. and Jakobsson, T. 2011. Successful discrete-event simulation
offshoring, Master of Science thesis, Göteborg, Sweden, Chalmers
University of Technology.
Awuah-Offei, K., Temeng, V. A. and Al-Hassan, S. 2003. Predicting equipment
requirements using SIMAN simulation – a case study, Mining Technology,
112, (3), 180–184.
Basu, A. J. and Baafi, E. Y. 1999. Discrete event simulation of mining systems:
current practice in Australia, International Journal of Surface Mining,
Reclamation and Environment, 13, (2), 79–84.
Ben-Awuah, E., Kalantari, S., Pourrahimian, Y. and Askari-Nasab, H. 2010.
Hierarchical mine production scheduling using discrete-event simulation,
International Journal of Mining and Mineral Engineering, 2, (2), 137–158.
Cain, P. 1999. Developments in coal pillar design at Smoky River Coal Ltd,
Proceedings, 2nd international workshop on coal pillar mechanics and
design, Alberta, Canada, 15–21.
Doe, D. C. and Griffin, W. F. 1987. Experimental design and mining system
simulation. Continuous surface mining, in Proceeding, 1st international
symposium on continuous surface mining, (eds. T. S. Golosinski and
F. G. Boehm), 317–324, Clausthall-Zellerfeld, Trans Tech Publications.
Dunrud, C. R. 1998. Engineering geology applied to the design and operation
of underground coal mines, Bulletin, US Geological Survey, Information
Services, 2147.
Farmer, I. 1992. Room and pillar mining, SME Mining Engineering Handbook,
2, 1681–1701.
Fishman, G. S. 2001. Discrete-event simulation: Modeling, programming,
and analysis. Series: Springer Series in Operations Research and Financial
Engineering.
Ghasemi, E. and Shahriar, K. 2012. A new coal pillars design method in order
to enhance safety of the retreat mining in room and pillar mines, Safety
Science, 50, (3), 579–585.
Hanson, B. D. and Selim, A. A. 1974. Probabilistic simulation of underground
production systems, SME Annual Meeting, SME Preprint No. 74AR32.
 Mining Technology  2016
Harrison, J. and Sturgul, J. R. 1988. GPSS computer simulation of equipment
requirements for the iron duke mine, 2nd large open pit mining conference,
AusIMM, 133–136.
InfoMine. 2013. Mine and mill equipment costs, UG-10–UG-48, Spokane
Valley, InfoMine USA Inc.
Kelton, D. W., Sadowski, R. P. and Sturrock, D. T. 2003. Simulation with
Arena, New York, McGraw-Hill.
Kiviat, P. J. 1969. Digital computer simulation: computer programming
languages, Santa Monica, CA, Rand Corporation, RM-5883-PR.
Kleijnen, J. P. C. 1998. Experimental design for sensitivity analysis,
optimization, and validation of simulation models, in Handbook of
simulation – principles, methodology, advances, applications, and practice,
(ed. J. Banks), 173–223, New York, Wiley.
Loui, J. P. and Sheorey, P. R. 2002. Estimation of non-effective width for
different panel shapes in room and pillar extraction, International Journal
of Rock Mechanics and Mining Sciences, 39, (1), 95–99.
McHaney, R. 1991. Computer simulation: a practical perspective, San Diego,
CA, Academic Press.
Michalakopoulos, T. N., Arvaniti, S. E. and Panagiotou, G. N. 2005. Simulation
of a continuous lignite excavation system, International symposium on
mine planning and Equipment selection (MPES 2005), Alberta, Canada,
1694–1706.
Nance, R. E. 1995. Simulation programming languages: an abridged history,
Proceedings, 27th Winter Simulation Conference, IEEE computer society,
1307–1313.
Pegden, C. D., Sadowski, R. P. and Shannon, R. E. 1995. Introduction to
simulation using SIMAN, New York, McGraw-Hill.
Pereira, S. P., Costa, J. F. C. L., Salvadoretti, P. and Koppe, J. C. 2010.
Simulação de produção em mina subterrânea de carvão com uso
de conjuntos mecanizados, Rem. Revista Escola de Minas, 63, (3),
581–589.
Petering, M. E. H. 2009. Effect of block width and storage yard layout on
marine container terminal performance, Transportation Research Part E:
Logistics and Transportation Review, 45, (4), 591–610.
Rice, S. V., Markowitz, H. M., Marjanski, A. and Bailey, S. M. 2005. The
SIMSCRIPT III programming language for modular object-oriented
Simulation, Proceedings, 37th winter simulation conference, 621–630.
Robson, T. A. 1984. The application of improved room-and-pillar techniques
at Smoky River Coal’s underground operations. SRCL Internal Report,
August.
Rockwell Automation, Inc. 2012. Arena simulation software. Rockwell
Automation [Online]. Available from: http://www.arenasimulation.com/
Arena_Home.aspx
Schriber, T. J. 1977. Introduction to simulation, in Proceedings of the 9th
conference on Winter Simulation-Volume 1 (p. 23). Winter Simulation
Conference, December, Gaithersburg.
Sturgul, J. R. and Harrison, J. 1987. Simulation models for surface mines,
International Journal of Surface Mining, Balkema, (I), 187–189.
Vagenas, N. 1999. Applications of discrete-event simulation in Canadian
mining operations in the nineties, International Journal of Surface Mining,
Reclamation and Environment, 13, (2), 77–78.
Van’t Ooster, A., Bontsema, J., van Henten, E. J. and Hemming, S. 2013.
Sensitivity analysis of a stochastic discrete event simulation model
of harvest operations in a static rose cultivation system, Biosystems
Engineering, 116, (4), 457–469.
Yuriy, G. and Vayenas, N. 2008. Discrete-event simulation of mine
equipment systems combined with a reliability assessment model based
on genetic algorithms, International Journal of Mining, Reclamation and
Environment, 22, (1), 70–83.
Zipf, R. K. 2001. Pillar design to prevent collapse of room and pillar mines, in
Underground mining methods: engineering fundamentals and international
case studies, (eds. W. A. Hustrulid and R. L. Bullock), 493–496, Colorado,
USA, SME.
9