Optimal sequencing and scheduling for a block/panel cave mining

Winston Rocher
M.Sc. Candidate, Delphos Mine Planning Laboratory, Advanced Mining Technology Center (AMTC),
Mining Engineering Department, University of Chile, CHILE. Address: Tupper 2069, Santiago, Chile.
Email: [email protected]

Milivoj Smoljanovic
M.Sc. Candidate, Delphos Mine Planning Laboratory, Advanced Mining Technology Center (AMTC),
Mining Engineering Department, University of Chile, CHILE. Address: Tupper 2069, Santiago, Chile.
Email: [email protected]

Enrique Rubio
PhD., Assistant Professor, Director at Delphos Mine Planning Laboratory, Advanced Mining
Technology Center (AMTC), Mining Engineering Department, University of Chile, CHILE. Address:
Tupper 2069, Santiago, Chile. Email: [email protected]

Nelson Morales
PhD., Principal Researcher at Delphos Mine Planning Laboratory, Advanced Mining Technology
Center (AMTC), Mining Engineering Department, University of Chile, CHILE. Address: Tupper 2069,
Santiago, Chile. Email: [email protected]

ABSTRACT

Currently, the mining industry creates the mine planning by using mathematical optimization. To
achieve this, a mathematical model is proposed with an objective function such as to maximize the
profit, to maximize the life of the mine, to minimize concentration of some pollutants, to minimize
mining costs, to get a determinate confidence level or to get ore reserves. Also, certain constraints
should be satisfied such as the production capacity rate or the plant production capacity rate. However,
a loss of the economic value can be appreciated in large underground mining projects either by using a
unique sequence of opening drawpoints (which is selected with a criteria, not necessarily related to the
optimization) or by not considering the construction time of the underground mine infrastructure. It
causes delays in the mine production with respect to the mine planning. Nowadays, in the mining
industry these are common practices.

If these considerations were incorporated into the optimization process, the solution would be different.
There would be another value for the objective function, and different opening drawpoint sequences.

This paper shows a novel, integrated way to make the mathematical optimization for underground
mines. It considers the opening drawpoints sequence and drift development as a result of the
optimization.

Finally, for production and development, the implementations of these models are respectively named
BOS2Undergound and UDESS. These models were already presented in APCOM 2011. This paper
corresponds to an actualization of that work.
 BIOGRAPHY

Winston Rocher Anda, a mining engineer and masters student at the University of Chile, has
participated in different activities and organizations during his studies, such as: scout group, student
council for mining engineering, IIMCh, SME and recently IMMS, he has been a finalist in the National
Math Olympiad, participate a study of the Large Magellanic Cloud for the European Southern
Observatory and had been an assistant for several courses in the university. All this culminated in the
award Juan Brüggen, for the best graduate of the mayor of mining engineering of the University of
Chile in 2010. Winston is currently pursuing a Mining Masters in the same university where he
completed his undergraduate degree, at Delphos Mine Planning Laboratory, working on underground
development sequencing and scheduling; finalizing his second undergraduate degree as a Civil
Engineer for a better understanding of solids and fluids; and, since April this year, working on the
project Codelco Chuquicamata Underground.
INTRODUCTION

Mine planning is the discipline of mining engineering that transforms a mineral resource into a mining
business. The outcome of this process is the production plan that delineates the best strategy to make
the mining business attractive. To achieve this purpose, mine planning incorporates the mathematical
optimization as a power tool with a lot of constraints and an objective function that represents the
strategy of the company or the client.

In particular, the underground mine planning is composed mainly of two components: production and
infrastructure preparation or development. Often, the production exercise is performed into two
decoupled stages: First, the mining exploitation unit sequence is determined (drawpoints or stopes as
required). Second, planning the development of the infrastructure as required. Although each one is an
optimal solution, to consider these two stages as independent processes does not necessarily conduce to
find the optimal solution. The last case is a local optimum, but a global optimum is required. The last
idea is as strange as imagining an open pit mine sequence without integrating the waste removal! This
can bring new and troubling problems that have already been documented in the industry (Díaz and
Morales, 2008).

In order to illustrate this decoupled process, Figure 1 shows six exploitation units with their respective
values (net profit) connected with paths or required developments to reach from one exploitation unit to
another. Each path has its cost to get to the next unit. Two different strategies have been proposed. The
first one is to start from the best value unit and then select the best value unit between the neighboring
connected values. The second strategy is to start from the path with the lowest cost, and select the
neighboring paths with the lowest cost until finished. In the example, the implementation of both
strategies results in different sequences or solutions with different net profits. However, both solutions
that extract the same mineral should be considered to solve the problem. Therefore, with this simple
example can be illustrated that this is not a trivial problem and deserves to be analyzed.

-­‐20   -­‐40  

20   80   40  
-­‐20   -­‐40   -­‐12   -­‐15   -­‐45  
20   80   40   Strategy  1   40   30   10
-­‐12   -­‐15   -­‐45   -­‐10   -­‐8  
40   30   10 Strategy  2   -­‐20   -­‐40  

-­‐10   -­‐8   20   80   40  
-­‐12   -­‐15   -­‐45  

40   30   10
-­‐10   -­‐8  
Figure 1. Conceptual example with six extraction units, where two strategies are shown.

The research summarized in this paper shows two novel and integrative tools for underground mining.
It considers the mine planning production optimization and the schedule of the infrastructure
development.

The first tool is called BOS2Underground, based on Blending Optimizer Sequencing and Scheduling
(BOS2). This tool incorporates the sequencing as an output rather than an input, and is subject to a
group of geometrical, precedence and logical constraint to generate the best feasible sequence. It also
considers the global mining system capacity and the capacity per mining system component. The
second tool is called Underground Development Sequencer and Scheduler (UDESS). This offers to
optimize the infrastructure development with a consideration towards mining production rates.
STATE OF THE ART

During the last twenty years optimization has been incorporated as a tool for mine planning. In this
sense, the latest and most important work will be reviewed to understand what has been researched.
Chanda (1990) used a computerized model for short-term production scheduling, combining simulation
with mixed integer programming. The problem with drawpoint scheduling for production was studied.
The goal was to reduce as much fluctuation as possible between periods of the average grade drawn.
This model considers geometric constraints between drawpoints. (For example: precedence).
Jawed (1993) used another model with the same objective function with a focus on operational
constraints, manpower requirements, extraction capacity, ventilation requirements, plant capacity, and
lower bounds on extraction quantity. The work was prepared for a room and pillars mine.
Trout (1995) presented a model to optimize the mine production schedule. He maximized NPV, and the
model was applied for sublevel stopping with backfill. The sequencing was considered, but geometric
constraints of sublevel stopping sequencing, permits more geometries than in panel or block caving
cases. This model is a good starting point; however, precedence constraints and capacities should be
modified. Also Carlyle and Eaves (2001) carried out a similar study for the Stillwater Mining
Company.
Rahal et al. (2003) used mixed integer linear programming for block caving to solve an optimization
problem. The deviation from the ideal draw profile was used as an objective function. Constraints of
capacity, precedence, material handling and maximum and minimum draw rate levels were considered.
Kuchta and Newman (2004, 2007) presented an optimization model to determine an operationally
feasible ore extraction sequence that minimizes deviations from planned production quantities.
Aggregation was used to optimize long-term production planning at an underground mine. The solution
was applied for a sublevel caving mine (Kiruna) with geometric precedencies defined in one direction,
horizontal (enough for sublevel caving, but not enough for block caving or sublevel stopping).
Rubio and Diering (2004) solved another optimization problem, by maximizing NPV for block caving.
Two slices were used to simulate columns in a discrete vertical model and the objective function of the
Rahal paper was tested. The model used precedence constraints, defined only for immediate neighbors.
Geometrical precedence, considering the period in which the predecessor drawpoints are mined, was
not analyzed.
Sarin and West-Hansen (2005) solved a planning optimization problem with mixed integer linear
programming. NPV maximizing was used as an objective function and penalties for deviations in
production and qualities were applied. The method was developed for room and pillar and longwall
mining systems. The model contained capacity, sequence (constraints for immediate neighbors) and
construction constraints.
Queyranne et al. (2008) presented a model for block caving that maximized NPV and used the capacity
constraints of mine production, the maximum opened and active drawpoints and the neighborhood to
continue production. Binary variables were considered and drawpoints can only be active for a
determined number or periods. The constraints of neighboring drawpoints do not consider a range of
time to mine them, but all neighbors have to be mined in the same period. However, the model does not
have a capacity constraint per drawpoint.
O’Sullivan et al. (2010) solved a long-term extraction and backfill of an underground mine by deciding
which and when to extract or backfill each area, depending on the global availability of material to
refill, as well as simple relations for geotechnical stability. By putting all the production tasks in a gantt
chart that generates more constraints then desired, a time heuristic was used to solve the large model.
The variable is binary and does not have control over the eventual idle times.
It is necessary to understand that over the last twenty years the underground mine planning research has
been focused on mineral extraction. The most difficult issue has been considering all the factors that
have to be rechecked in the caving case.

THE MODELS
1. BOS2Underground

This model attempts to incorporate sequencing and capacity constraints, locating the sequence in time.
The model is based on BOS2 (Vargas et al., 2009), which was developed in Delphos Laboratory for
open pit mining. The constraints in the model were mining and processing capacities, geometalurgy,
stocks, and geometric constraints. It was necessary to adapt the model for underground caving mining
and adding some new constraints.
BOS2U includes an objective function and different constraints that consider precedence, logical
relations, global capacity, capacity per components (shaft and crusher), minimum AL mined neighbors
required to mine a block and a maximum advance per period to control dilution. All these topics and
the details of the model have been published recently (Smoljanovic, 2011).

Figure 2. Objective function, variables and constraints of the BOS2U model.

The problem is summarized in Figure 2. In particular, the constraints minimum neighbors and
maximum advance induces the sequence and they permit that the sequence corresponds to a panel
caving method. “Minimum neighbors” is a constraint that considers a neighbors subset C of block B.
To extract B, all blocks contained in C should be mined in time ∆. Maximum advance is a constraint
that permits to extract neighbors until determined extent, considering the perpendicular axis to the front
cave line in a period that permits to control the dilution and give an opening continuity for production.

2. UDESS

This model was developed in Delphos Laboratory in 2011 as a new line of research that tries to solve
the mine development problem by transforming all the “segments” (related to drift or infrastructure) of
a mine layout in activities that have the necessary attributes to prepare a model to be optimized. The
attributes of the activities are a maximum and minimum development rate, a cost or profit for this
activity, required resources to get the maximum and minimum development rate (equipment, materials,
air ventilation, workers, etc.) and physical and operational precedence. Physical precedence orders the
activities in the space according to the mine access, while operational precedence orders the activities
by a logical criterion according to the associated processes (for example the crusher to draw mineral
shown in Figure 3).
 Figure 3. Physical and Operational Precedence.

The model considers maximizing the net profit as the objective function by implementing continuous
variables and constraints related with the precedence, with the available resources to be distributed in
all the activities, global capacity of mine capacity in plants and paths. Path constraints try to add the
time consumption on each activity on every period so idle times do not appear. Moreover, to solve the
larger problem it was necessary to make a time heuristic to get to a final schedule with small time
periods over the entire lifetime of the mine. All the details and explanations of the model have already
been published in APCOM 2011 (Rocher, 2011).
Figure 4 shows the cycle to achieve the optimized solution. It is important to note that the step
involving the transformation of a CAD layout to activities was done using commercial design software,
(in this case MINE2-4D).
While the text file introduced to UDESS is optimized, one of the result tables can be reintroduced in
BOS2U and another can be reincorporated to MINE2-4D. This will insure the overall process considers
the two main components of mine planning and gives a global optimum.

Figure 4. UDESS optimization cycle.

ANALYSIS WITH BOS2Underground

The case study is a mine that has a production rate of 30.000 tons per day (tpd). The layout has 332
drawpoints. The considered exploitation unit will be a drawpoint with the idea to model a panel caving
method. Currently the sequence has already been developed, but the idea is to show how the results
vary when the mining system is changed. Production is delivered to 4 crushers located on the ore body
footwall. Figure 5 shows a three dimensional view of the mining system, which has four, 10.000 tpd
crushers and one load-hawl-dump (LHD) per cross-cut. The last system is the real case, and the layout
was tested with different mining systems. The systems are classified by crusher capacity in three
groups: Group E: 4, 5, and 6 crushers (10.000 tpd each one), group F :10, 11 and 12 Chutes to a big
crusher (5.000 tpd each chute) and group G: 3, 4 and 5 crushers (12.500 tpd each one).
 Figure 5. Case study mining system.

1. Inputs at BOS2U and considered Mining System.

The following table (Table 1) shows the required inputs to solve the model:
Table 1. Principal static parameters used in the model.
Value unit
Production 30 ktpd
Planning horizon 13 periods
Discount rate NPV 10 %
Maximum advance 5 dpt/period
Minimum neighbors 6 dpt
Periods to mine the minimum neighbors 2 periods
Periods to mine a drawpoint. 3 periods
Minimum exploitation per column 30 %
Capacity per drawpoint 80 ktpy
Capacity per crosscut 3,000 tpd

2. Results.

Figure 6 (left) shows the best production plan in each defined group with a clear plateau which lasts six
years, considering a horizon planning of 13 years. The grades decrease because the optimization was
performed with NPV as the objective function. The mining system reaches the production in the three
cases perfectly.

12   1.2   Produc6vity  per  crosscut  

Ton  E.3  
10   1   [Mton]  
3.3  
Grade  Cu  [%]  

Produc6vity  [Ktpd]  

8   0.8  
Ton  F.3   2.8   Case  
Tonnage  [Mton]  

6   0.6   [Mton]   E.3  

2.3  
Case  
4   0.4   Ton  G.3   1.8   F.3  
[Mton]   Case  
2   0.2   1.3  
G.3  
Cu  
0   0   0.8  
Grade  E.
0   5   10   15   1   3   5   7   9   11   13   15   17   19  
3    [%]  
Years   Cross-­‐cut  number    
Figure 6. Production plan for the considered mining system.

Figure 6 (right) shows cross-cut performance. In relation to the maximum production per LHD per
cross-cut not all the LHDs reached the capacity. On the other hand, in the case of the crushers the
capacity was reached in most of them. The last idea indicates an imbalance between LHD and
crushers, so these components should be balanced to obtain the best performance. Case F presents the
worst production performance; due to this case this mining system has more constraints than the others.
The relation between chutes and cross-cut is stricter. However the curves shown are regular enough to
accept the behavior as regular.
Table 2. NPV results best and worst case per group
Case E F G
Best [MU$] 1,123 1,105 1,124
Worst [MU$] 1,088 1,110 1,088
Difference [MU$] 35 5 36
Perc. Diff. [%] 3.1 0.5 3.2

Finally the NPV results per group are shown in table 2. Important differences can be noted between
group F and the other two groups. Firstly, in case F the lowest value, t, is the best value, but the gap
between the worst and best value for case F is the lowest. This is due to the fact that the three cases in
group F are very similar. Secondly, values in the cases E and G are similar, but with a different number
of crushers. Note that the NPV is directly related with sequence, so if the system changes the sequence
changes.

ANALYSIS WITH UDESS

1. Testing the hypothesis of influence of the mining development in the production plan

106 activities were run with a time horizon of 60 monthly periods of a sublevel open stoping (SLOS)
mine. The 106 activities included the drifting development, stopes drilling and extraction of the ore to
complete all the mine operations. Within the activities, the equipment considered is a Jumbo drill for
drifting, a Simba drill for radial drilling and a LHD for ore production.

The orebody consists of three vertical and parallel veins with each one having three levels of stopes.
The main access is a ramp that connects the first ore body. From this ramp, drilling and production
galleries are constructed for each one of the three levels mentioned. Each level (in each ore vein)
contains three stopes that are extracted with a different profit and unclear trend. A set of galleries are
connected from the first to second orebody, which tries to replicate the infrastructure exposed in the
first case with the same characteristics and number of stopes. Eventually, the infrastructure connects
with the third body in the same system to complete the entire mine. (Figure 7).

Figure 7.Sublevel open stoping layout

 
a. Results varying the available resources.

When the available resources for the project change, such as equipment and material for support, the
only effect would be that the whole mine increases its life. However, this does not change the order of
the activities. Since the resources are completely exhausted, the sequence of production changes its
geometry and main direction. This is better, but not always evident.
b. Results varying the drifting speed.

This test consists of putting a variation of drifting speed for development activities. As shown in table
3, with some rates there is a common sequence (40 and 70 percent of speed). When extracting the first
stopes level of each body and then move deeper, preferring to develop the horizontal galleries and at the
end the decline ramps at first. By contrast with a higher drifting speed than stated before, drastic
changes appear, preferring to extract all the ore from a particular vein and then develop the drift to
connect the mine with the following vein due to higher net profit. Finally, with an even greater drifting
speed, it is possible to select specifically which stopes will return higher values and eventually get a
decreasing grade profile, so the geometry and sequence of opening stopes changes.
Table 3. Variation in the order of exploitation of stopes changing the drifting speed.
Drifting Speed
Level
40% 70% 100% 130% 160%
HIGH 3 1 1 1 1 24
HIGH 2 2 2 2 2 1
HIGH 1 3 3 3 3 2
MED 3 10 9 7 4 5
Body 1

MED 2 11 10 8 5 6
MED 1 16 26 22 6 26
LOW 3 26 16 20 7 19
LOW 2 18 17 12 8 9
LOW 1 19 18 13 9 10
HIGH 3 7 6 27 10 27
HIGH 2 4 4 4 11 3
HIGH 1 8 8 23 12 20
MED 3 12 11 9 13 7
Body 2

MED 2 15 14 11 14 16
MED 1 17 15 25 15 21
LOW 3 24 24 19 16 17
LOW 2 20 19 14 17 11
LOW 1 23 21 16 18 14
HIGH 3 9 7 6 19 23
HIGH 2 5 5 5 20 4
HIGH 1 6 25 26 21 22
MED 3 13 12 10 22 8
Body 3

MED 2 14 13 18 23 15
MED 1 27 27 24 24 25
LOW 3 25 23 21 25 18
LOW 2 21 20 15 26 12
LOW 1 22 22 17 27 13

2. Panel/Block caving mine with a defined production plan

1454 activities were run with a time horizon of 30, six-monthly periods of caving mine. The
exploitation units are 332 drawpoints spaced around the entire production level. The available
controlled resource has the following required equipment: Jumbo, Simba, LHD, shotcrete machine,
Bolter Machine and Raise Borer Machine.

The purpose is to start from a defined production plan and optimize the entire project. The first
evaluation verifies if the plan is feasible with the given parameters. The next evaluation determines if
the plan has some flexibility to search the best option of extraction given the mining design layout.

Figure 8 shows the obtained production plan considering these inputs and varying the drifting speed.
The production plan has little perturbation considering differences between the cases. In the case of 80
percent of the original maximum advance rate, it has to start a year later than the basic case. In the case
of 20 percent of the original maximum advance rate, the mine can be opened more quickly. In general,
this model has more flexibility and it can make some changes to get a better total value.
 Figure 8. Resulting production plan from the optimization varying the drifting speed.

All these cases have some differences of how much tonnage extract from each drawpoint in each time.
The majority of the cases keep the same sequence, excluding the case of 65 percent of the original
maximum advance rate. In this case the ore is not drawn equitably in all sectors.

CONCLUSIONS

In the case of BOS2U the material handling system has important effects that define the results. If the
mine planer does not incorporate this, the sequence could be completely different to the obtained with
the material handling system considered, because the developed plan does not necessarily respect the
constraints. The mine planner should test different material handling systems with accordance to the
real possibilities. The options should not be too overwhelming, but to select the best system it has an
important retribution in the final results (in value, depending on the objective function).

In the case of UDESS the optimization was reached incorporating connections and accessibilities. The
proposed tool could be used to determine bottle necks, critical paths and infeasible solutions in a short
time with a lot of variables and constraints. UDESS considers the sequence of the set of exploitation
units connected as an output.

The opening sequence is in accordance with the objective function. This sequence is a perfect reference
or guide for mine planner to have the best strategy for the mine with the possibility to do different tests
in a short time. The sequence should always be a result of the set of conditions. (For the two models)

The best result will be reached when production and infrastructure present a congruence to solve the
optimization problem. The production plan could be very different if the mining development is
considered. This paper presents two potent tools that, if combined properly in an engineering cycle, will
obtain a feasible and robust solution. The solution considers the interaction between the two mentioned
strategies, generating a realistic plan.

ACKNOWLEDGEMENTS

The authors would like to thank Delphos Mine Planning Laboratory, University of Chile and
CONICYT for providing the tools and funds that made this research possible.
REFERENCES

Carlyle, W M and Eaves, B C, (2001). Underground planning at Stillwater Mining Company,

Interfaces, 31(4):50-60.

Chanda, E. K. C. (1990). An application of integer programming and simulation to production planning

for a stratiform ore body. Mining Sci. Tech. 11(2) 165–172.
Díaz, G and Morales, E, (2008). Tunneling and construction for 140,000 tonnes per day – El Teniente
mine – Codelco Chile, in Proceedings MassMin 2008, pp. 83-96 (Luleå University of Technology:
Luleå).

Jawed, M. (1993). Optimal production planning in underground coal mines through goal programming:
A case study from an Indian mine, in Proc. 24th Internat. Sympos. Appl. Comput. Oper. Res. Mineral
Indust. (APCOM), pp. 44-50 (CIM: Montréal).

Kuchta, M, Newman, A and Topal, E, (2004). Implementing a production schedule at LKAB’s Kiruna
Mine, Interfaces, 34(2):124-134.

Newman, A and Kuchta, M, (2007). Using aggregation to optimise long-term production planning at an
underground mine [online]. Available from: <http://inside.mines.edu/~anewman/> [Accessed:
January, 2010].

O’Sullivan, D and Newman, A, (2010). Long-Term Extraction and Backfill in a Complex Underground
Mine, in Proc. 4th Internat. Confer. Mining Innov. (MININ), pp. 245-250 (Gecamin: Santiago).

Queyranne, M, Parkinson, A, McCormick, S T, Diering, T, Malkin, P and Wolsey, L, (2008). The

drawpoint scheduling approach to production planning in a block cave mine, in Proc. Workshop
Operation Research in Mining, Viña del Mar, Chile, pp 31-38.

Rahal, D and Smith, M, (2003). The use of mixed integer linear programming for long term scheduling
in block cave caving mines, in Proc. 31th Internat. Symps. Appl. Comput. Oper. Res. Mineral Indust.
(APCOM), pp. 123-131 (SAIMM: Cape Town).

Rocher, W, Rubio, E and Morales N, (2011). Eight-Dimensional Planning – Construction of an

Integrated Model for Mine Planning Involving Constructability, in Proc. 35th Internat. Sympos. Appl.
Comput. Oper. Res. Mineral Indust. (APCOM), pp. 393-406 (AusIMM: Wollongong).

Rubio, E and Diering, T, 2004. Block Cave production planning using operation research tools, in
Proceedings MassMin 2004, pp. 141-149 (IIMCh: Santiago).

Sarin, S C and West-Hansen, J, (2005). The long-term mine production scheduling problem, IIE Trans,
37(2):109-121.

Smoljanovic, M, Rubio, E and Morales N, (2011). Panel Caving Scheduling Under Precedence
Constraints Considering Mining System, in Proc. 35th Internat. Sympos. Appl. Comput. Oper. Res.
Mineral Indust. (APCOM), pp. 407-417 (AusIMM: Wollongong).

Trout, L, (1995). Underground mine production scheduling using mixed integer programming, in Proc.
25th Internat. Sympos. Appl. Comput. Oper. Res. Mineral Indust. (APCOM), pp 395-400 (AusIMM:
Melbourne).

Vargas, M, Morales, N, Rubio, E and Mora, P, (2009). BOS2, Sequencing model to mine reserves, in
Proc. Mine Planning 2009 (Gecamin: Santiago).