Você está na página 1de 7

JOURNAL OF COAL SCIENCE & ENGINEERING (CHINA) pp 113119 DOI 10.1007/s12404-010-0201-y Vol.16 No.

2 June 2010

A new algorithm for stope boundary optimization


TOPAL Erkan1, SENS Jeroen2,3
( 1. Western Australia School of Mines, Curtin University of Technology, Kalgoorlie of Queensland, Brisbane WA 6433, Australia; 2. The University 2628CJ, the Netherlands ) QLD 4072, Australia; 3. Delft University of Technology, Delft

The Editorial Office of Journal of Coal Science and Engineering (China) and Springer-Verlag Berlin Heidelberg 2010

Abstract Stopes can be simply defined as an underground opening from which ore has been excavated. Selection of the best combination of available stope boundary will directly affect the profitability of the operation. While a few attempts has been initiated to generate the optimum stope boundary for underground mining, they fail to guarantee a true optimality in three-dimension block models. This paper proposed a new methodology which can find optimum stope layout for a given resource model in three-dimensions. The paper initially critically reviewed important stope boundary optimisation studies thus far, then proposed a new methodology in order to find the best stope layout for a given deposit. Subsequently it applied the proposed methodology into a block model to test its ability of producing optimum results and demonstrated its applicability in a number of different scenarios. In the last section, further analysis on strategies to find the optimum stope boundaries were demonstrated. The results prove that the proposed algorithm can find optimum stope boundaries and layouts in three-dimension for different stope sizes and stope selection strategies. Keywords Stope design, optimisation, underground mining, stope layout algorithm

Introduction
Mine optimisation techniques have been applied since early 1960s to a variety of mining problems. A variety of algorithms and methodologies have revolutionised surface mine design and schedule. However, underground mine optimisation has attracted more attention in the last 10 to 15 years with a focuse on three main areas: optimisation of development and infrastructure, Optimisation of stope boundaries and optimisation of production schedule using predefined stope boundaries (Topal, 2008). The researches on development optimisation mainly focused on cost minimisation. In other words, it optimises a cost function over space of feasible solutions in underground. The proposed analytical solution for the problem is based on network optimisation
Received: 17 December 2009 E-mail: e.topal@curtin.edu.au

model. In this model, the given draw points and surface portal correspond to fixed nodes of the network and known as terminals. The ramp in the mine represented by links in the network and junctions at which three or more ramps meet are represented by variable nodes in the model and known as Steiner points. Construction and haulage costs of each link in the network are modelled and solved. Other than main assumptions within the model, the other main short comings of the research so far is not being able to consider the grade distributions of the orebody and not being able to put the development in high grade ore zone first for net present value (NPV) maximization (Alford, 1995; Brazil et al., 2003; Brazil and Grossman, 2003). Limited amount of studies carry out on stope scheduling. Trout formulated a Mixed-Integer Programming (MIP) model for scheduling sublevel stop-

114

Journal of Coal Science & Engineering (China)

ping operation. The model was applied to a representative data set which came from Mount Isa, containing 55 stopes. Although the generated schedule was not optimal, it showed some merit of employing MIP techniques over manual techniques for improved scheduling and sequencing of stopes (Trout, 1995). Nehring and Topal advanced the model in (Trout, 1995) by formulating a new constraint to limit multiple fillmass exposures without violating other constraints to make the model more applicable. Small case study demonstrates the benefits of using MIP for generating production schedules over a common manual method of selecting production from the next highest available cash flow stope (Nehring and Topal, 2007). Little et al. revised the same model in order to reduce the number of variables and thereby reduce the solution time. The proposed model is able to reduce the number of variables by 80% by utilising natural sequencing and natural commencement between phases. The proposed model allow to solve the scheduling problem for large scale stopping operations (Little et al., 2008). While a number of studies have been conducted on the underground development and stope scheduling optimisation, this paper primarily focuses on stope boundary optimisation. Alford has developed a stope optimising method which is called floating stope. The floating stope algorithm helps determining the boundaries for minable ore and is implemented in the mine design package Datamine. In the algorithm, a stope with the minimum stope dimensions is floated throughout the orebody, the grade inside the stope is calculated by averaging the grades of the blocks it contains, all the stopes having a grade higher than the cut-off grade are selected. It creates two envelopes, the minimum envelope is generated by taking the union of all best grade stope positions for every block. The maximum envelope is the combination of all possible stope positions that have a grade higher than the cut-off grade. The optimum stope mine design is located in between these two envelopes. The major problem of the floating stope algorithm is overlapping stope and it occurs when two or more stopes share one or more high-grade blocks. In some occasions each stopes may individually economical but the overall economical envelope may not (Alford, 1995; Alford et al., 2007). In order to help overcome the shortcomings of the floating stope algorithm, the multiple pass floating stope process (MPSFP) was developed. The process relies on the same principals as the floating stope al-

gorithm. It is basically delineate economic ore zones using multiple sets of input parameters i.e., head grade, cut-off grade, maximum waste. The generated envelopes based on input parameters can provide extra information during the mine design process helping the designers to improve the efficiency of their mine design. Although the method can assist in stope boundary selection and design, it does not generate optimum stope layouts (Cawrse, 2007). Alford and Hall developed a new tool for automated stope design which generates set of possible stope shapes for assigned cut-off grade ranges. The new approach incorporates three stages: first the economic zones are identified by sampling the mineralisation on a regular grid, and the second stage uses the technique stope shape annealing to adjust the shape to a local maximum, the third stage generates stoping blocks or stope groups, it takes into account the location of stope shapes, leaving out all stopes that cannot be mined or are inaccessible. The tool seems to automate the design process whether it adds any real optimisation to the design process can be doubted and also needs further enhancement, such as mining cost to be a function of the size and shape of the stope, evaluation of the contributions of individual ore type to stope value (Alford and Hall, 2009). Ataee-pour introduced heuristic approach called Maximum Neighbourhood Value(MNV) for stope boundary optimisation. The method works in a similar fashion to the floating stope but it uses different approach to define the envelope. The optimiser starts looking for a stope around the first block in the block model which is located in one of the corners of the block model. After analysing the stopes that contain this block it chooses a stope, tags the blocks and moves to the next block. In this way, the algorithm will systematically move through the block model. Because the algorithm does not allow partial blocks to be included in a stope, the stope size has to be rounded to be able to contain a positive integer number of stopes. Although the algorithm provides a theoretical basis for a stope boundary optimisation algorithm, it does contain a several fundamental shortcomings. The first shortcoming is that varying the starting point for the algorithm can provide various sets of stopes for the orebody, proving that it does not always find the optimum mine design. Also the algorithm uses a static economical block model which does not allow stope costing factors to be taken into account. The size of the stope will determine what the mining costs are

TOPAL Erkan, et al. A new algorithm for stope boundary optimization

115

therefore it is not sufficient to assume a constant mining cost over the whole orebody (Ataee-pour, 2004; Ataee-pour, 2005). Dimitrakopoulos and Grieco developed a new probabilistic optimisation approach, only stope design method that acknowledge uncertainty, based on MIP. The orebody subdivided in layers, panels, and rings. Each panel comprises of a series of rings, each layer comprises of a number of panels and orebody comprises a number of layer. The objective function of the model maximises the mined metal content at a given time. The constraints limit the minimum and maximum mining rings as well as the size of the pillars to be left unmined between two primary stopes. The methodology has been applied to data from Kidd Creek Mine, copper-zinc-silver deposit, Ont., Canada. Sequential Gaussian simulation method was used to generate 40 realisations on a 1.5 m by 1.5 m by 1.5 m grid. Simulated realisation are reblocked minable rings with a block size of 15 m by 20 m by 40m. The orebody has two layers each with eight potential stoping panels which each contain 18 rings. Five stopes and pillar design are created and compared for copper grade, ore tonnes and economic potential using risk profiles. Although the proposed method greatly assists stope design by considering geological uncertainty and it is the first stope design methodology that has acknowledged uncertainty, it has potential drawbacks. Because the methodology based on rings that have been predefined in terms of location and size to determine the most profitable stopes, it will find optimum stopes layout based on those rings. This methodology does not allow accurate examination of the orebody over smaller areas and examination of stopes in different locations. Moreover, as each ring is represented as a binary variable within the MIP model, this encourages a very long solution time as the number of ring increases within the model (Grieco and Dimitrakopoulos, 2007; Dimitrakopoulos and Grieco, 2009). Ovanic and Young introduced a mixed integer programming model (MIP) to optimise stope boundaries. The model locates the optimal starting and ending locations for mining within a row (mining panel) in a block model and in this way establishes the optimum stope boundaries. To determine the optimum starting and ending location of each panel, two piece-wise linear cumulative functions are calculated. The advantage of this technique compared to the others noted is that the block geometry is not required to be regular or orthogonal. Its disadvantage is, however, that the algo-

rithm optimises the stope boundary along the row of blocks in only one-dimension. Examples for using the algorithm in three dimensions do not exist. Also the algorithm only allows the user to optimise the design of extraction rows only, as the locations of these rows are determined earlier in the design process the algorithm only partially optimises the mine design (Ovanic and Young, 1995; Ovanic and Young, 1999). In the next few sections, a new methodology in order to find the best stope layout for a given deposit is detailed. Then, an application of the proposed methodology to a real block model is demonstrated with a number of different scenarios. Further analysis on strategies to find the optimum stope boundaries are demonstrated. Comments and conclusion follow.

Proposed stope boundary optimisation methodology

The proposed stope boundary and layout optimisation methodology consists of 3 basic elements. These are block converter, stope optimizer and visualiser. 1.1 Block converter In order for the stope boundary and layout optimiser to work, the block model that is used can only contain blocks with same size. Therefore, a block converter is created which can convert a block model containing blocks with multiple sizes into a block model containing blocks with only one size. It also allows blocks of each possible size to be created, thus enabling stopes of each possible size to be generated. The block converter works by analysing which blocks and which part and volume of the block are positioned inside the new block. It then calculates the values of the new block by taking the product of the value of each block inside the new block and the volume of the part that is located inside the new block. These values are then summed and divided by the total volume of the new block. It does this for all the blocks in the new block model (Sens, 2008; Sens and Topal, 2008). 1.2 Stope boundary optimiser The optimiser uses the block model generated by the block converter. The input parameters include the range of all the possible stope sizes (or the stope sizes that have to be used), ore price(s) per tonne sold material, mining and processing costs per tonne of mined material, backfill costs per cubic meter as well as a fixed stope start-up costs. From these values, the algorithm determines the optimum stope boundaries and

116

Journal of Coal Science & Engineering (China)

layout for a given orebody. The stope boundary optimiser was made using Matlab and performs the following actions. The program first loads the block model with uniform block sizes which is generated by the block converter. After the block model is loaded the optimiser then works through the following process: Step 1: Optimiser starts plotting starting from the smallest available stope size on every possible location. Of all the stopes with positive profit their location and profit are noted down in the stope list. The next largest stope is then chosen and the same actions are performed. This is repeated until all the different stope sizes have been analysed. A table is formed containing all the possible stopes and their profits. Step 2: An envelope is being created on every individual stope. The values of all the stopes within this envelope are summed and divided by the number of stopes inside the envelope. This value which is the average envelope stope profit is also noted down in the stope table for each possible positive stope. Step 3: Stopes are then selected based on different criteria which are: (1) The stope profit. (2) The stope profit per square metre. (3) The stope profit divided by its total mining time (in order to do this an extra calculation is carried out that estimates the mining time of the stope). The highest profit stope is first selected by noting its profit, location and size in the final stope layout table. The algorithm checks if the selected stope also has the highest average envelope stope profit from Step 2. If it is then blocks within this stope shell are tagged and the stope and all other stopes sharing common blocks are deleted from the stope table. Step 4: Step 3 is repeated to find the next highest profit stope. This process continues until all stopes from the stope table are removed. After this the final stope layout table contains only ranked positive non-overlapping stopes chosen in the order of the users preference. By selecting the stopes based on their average envelope stope profit it is guaranteed that the combination of stopes that is being selected by the optimiser is the optimum selection possible, provided that the envelope has sufficient size. 1.3 Stope visualiser The stope visualiser is a program that generates a three-dimensional view of the final stope layout from the optimiser. The profit from the stopes is indicated

with a colour and legend linking profit to its respective colour is provided.

Implementation of the proposed algorithm

The proposed algorithm allows stope boundaries and layouts to be generated according to a number of different stope design strategies. The optimiser can: (1) Generate optimum stope boundaries using one given fixed stope size. (2) Generate optimum stope boundaries using a range of given stope sizes. (3) Find the optimum stope size and generate the stope mine layout based on that stope size. (4) Make a stope design starting with stopes having the highest profit per cubic meter. The block model used contains 40 000 blocks which have a maximum dimension of 20 m20 m 20 m, and is illustrated in Fig.1.

Fig.1

Visualisation of test block model

Using this block model, four different cases were generated with the proposed optimiser, in all the cases the following costs were used: (1) Mining and processing cost: 45 $/t. (2) Backfilling cost of 10 $/m3. (3) Stope start-up cost: $80 000/stope. In Case 1, fix stope size of 40 m40 m40 m and a commodity price of 5 000 $/t was used. Then the stope size let to flow in a range of 20 m20 m20 m to 40 m40 m40 m in Case 2 with same commodity price as Case 1. For Case 1, the optimiser identified 45 000 possible stopes and of these, 307 were identified as having a positive value. From these positive stopes, 34 stopes were selected for the final stope mine layout. This layout produced a revenue of $265 129 466. The proposed design can be seen in Fig.2(a). For Case 2, the optimiser identified 360 000 possible stopes and of these, 4 570 were identified as having a positive value. From these positive stopes, 180 stopes were selected for the final stope mine layout. This layout produced a revenue of $995 148 686. The proposed design can be seen in Fig.2(b).

TOPAL Erkan, et al. A new algorithm for stope boundary optimization

117

Fig.2

Fix and floating stope sizes with 5 000 $/t commodity price

Fig.3 Floating stope sizes with 4 000 $/t and 5 000 $/t commodity price

When comparing Case 1 and Case 2, it can be seen that Case 2 generates a larger overall profit, although the layouts were designed with the same commodity price. Because the optimiser has a stope size range to choose from, it has more options and therefore is capable of finding a more profitable mine layout. Case 3 and 4 have been generated with same input as Case 2 with an exception of commodity price. Case 3 and 4 was based on 4 000 and 3 000 $/t commodity prices respectively. For Case 3, the optimiser identified 360 000 possible stopes and of these, 2 816 were identified as having a positive value. From these positive stopes, 145 stopes were selected for the final stope mine layout. This layout produced a revenue of $644 745 013. The proposed design can be seen in Fig.3(a). For Case 4, the optimiser identified 360 000 possible stopes and of these, 2 615 were identified as having a positive value. From these positive stopes, 130 stopes were selected for the final stope mine layout. This layout produced a revenue of $329 010 756. The proposed design can be seen in Fig.3(b). When comparing Case 2, 3 and 4, it can be seen that higher commodity price, the larger the number of stopes and profits.

Optimum stope boundary selection strategies

The optimum stope boundary can be selected based on the profit generated by individual stopes. This means that the stope with the highest possible profit will be selected first. Another method that can be used to select the optimum stope boundary is based on the highest profit per time unit (stope profit/total production time). This gives a profit per time unit factor for each stope. Again, the stope with the highest factor is selected first and so on. In order to calculate the profit per time unit the stope mining time has to be calculated. For the following examples the stope mining time is calculated based on two assumptions; all stopes have a fixed start-up time, the production and backfill time have a linear relation with the stope volume. The calculation of NPV is based on the mining of single stope at a given time. Both strategies have been tested with the following example with respect to profit and NPV values. In this example it was chosen to give a production time which has a linear relationship with the stope volume. All other variables (i.e. the block model, mining cost, start-up cost, start-up time, production time, stope

118

Journal of Coal Science & Engineering (China)

range) were kept constant for both cases. A table was generated containing the profit from each selected stope using both strategies. The NPV for both strategies were calculated based on the ranking of these stopes and mining these stopes one at a time without considering any geotechnical constraints. The layout for Strategy 1, where selection was made based on the highest profit per stope, can be seen in Fig.4(a). The layout for Strategy 2, where selection was made based on the highest production value per time unit, is presented in Fig.4(b).

costs. Strategy 2 however provided a higher NPV. Thus selecting stopes based on their profits per time unit can result in a higher overall NPV. The comparison of generated NPVs and the profits from both strategies can be seen in Fig.5.

Fig.5 Profit and NPV comparison for overall profit and profit per time unit strategies

Conclusions

Fig.4

Stope mine layout with the highest profit per stope and the highest production value

The results from the two optimisation strategies are presented in Table 1.
Table 1 Results of highest profit per stope and high production value per time unit results
Strategy 1 24.9 years $42 614 151 $21 669 876 Strategy 2 24.6 years $41 390 076 $22 380 522

This paper has presented a new optimisation approach to stope layout that accounts for variable stope sizes, block values and the cost of mining in three dimensions. The major advantage of this method over existing methods is that it can find true optimum stope layout in a three dimensional block model. Furthermore, it allows to optimise single as well as variable stope sizes. The application of the proposed approach at a real case block model demonstrated how it can evaluate and find optimum stope layout with different stope sizes and commodity prices. The last section of the paper evaluates the starategies that can be applied to stope boundry definition. Two different strategies have been applied to stope boundary optimisation. Although the strategy based on highest profit per stope can generate a better overall profit, the strategy of highest profit per time unit provides a better NPV. The method presented herein can be further advanced by integrating development and infrastructure optimisation, stope boundary optimisation and stope production scheduling optimisation into one model. References
Alford C, 1995. Optimisation in underground mine design, 25th International APCOM Symposium, Melbourne. The Australian Institute of Mining and Metallurgy, 213-218. Alford C, Brazil M, Lee D H, 2007. Optimisation in underground mining, in Handbook of Operations Research in Natural Resources, 561-577. Alford C, Hall B, 2009. Stope optimisation tools for the selection of optimum cut-off grade in underground mine design,

Projects Mine life Profit NPV

As expected, Strategy 1 provided a higher profit. This is because the average size of the selected stopes is larger which results in comparatively lower start-up

TOPAL Erkan, et al. A new algorithm for stope boundary optimization Proceedings of project evaluation conference, Melbourne, 137-144. Ataee-pour M, 2004. Optimisation of stope limits using a heuristic approach. Mining Technology, 113:123-128. Ataee-pour M, 2005. A critical survey of the existing stope layout optimization techniques, Journal of Mining Science, 41: 447-466. Brazil M, Grossman P A, 2003. Grossman, access layout optimisation for underground mines, Australian Mining Technology Conference, Queensland, Australia, 119-128. Brazil M, Lee D H, Van Leuven M, Rubinstein J H, Thomas D A, Wormald N C, 2003. Optimising declines in underground mines, mining technology. Trans. of the Institution of Mining and Metallurgy, Section A, 112: 164-170. Cawrse I, 2007. Multiple pass floating stope process, Strategic Mine Planning Conference. Perth, Australia, 87-94. Dimitrakopoulos R, Grieco N, 2009. Stope design and geological uncertainty: quantification of risk in conventional designs and probabilistic alternative, Journal of Mining Science, 45(2): 152-163. Grieco N, Dimitrakopoulos R, 2007. Managing grade risk in stope design optimisation: probabilistic mathematical programming modal and application in sublevel stoping, Mining technology, 116: 49-57. Little J, Nehring M, Topal E, 2008. A new mixed integer programming model for mine production scheduling of sublevel stopping operation, crc mining australian mining

119

Technology Conference. Queensland, Australia. Nehring M, Topal E, 2007. Mixed integer programming model for mine production scheduling of sublevel stoping Operation, CRC Mining Australian Mining Technology Conference. Western Australia. Ovanic J, Young D S, 1995. Economic optimization of stope geometry using separable programming with special branch and bound technique, 3rd Canadian conference on Computer Applications in the Minerals Industry. 129-135. Ovanic J, Young D S, 1999. Economic optimization of open stope geometry, 28th international APCOM symposium. Colorado school of Mines, Golden, Colorado. USA, 855-862. Sens J J, 2008. Stope boundary optimisation, BSC thesis, University of Queensland, Brisbane, Australia. Sens J J, Topal E, 2008. A new algorithm for stope boundary optimisation, The AusIMM New Leaders' Conference. Brisbane, Australia, 25-28. Topal E, 2008. Early start and late start algorithms to improve the solution time for long term underground mine scheduling, SAIMM, South African Institute of Mining and Metallurgy Journal, 108(2): 99-1 079. Trout L P, 1995. Underground mine production scheduling using mixed integer programming, 25th International APCOM Symposium, Melbourne. The Australian Institute of Mining and Metallurgy, 395-400.

Você também pode gostar