Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

download Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

of 6

Transcript of Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

  • 8/22/2019 Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

    1/6

    Development and Application of Whittle Multi-Mine at Geita Gold

    Mine, Tanzania

    T Joukoff1, D Purdey2 and C Wharton3

    ABSTRACT

    In the past, life of mine scheduling at Geita Gold Mine, Tanzania, hasbeen a largely manual process involving the optimisation and scheduling

    of each mine as a separate entity. The scheduling has been atime-consuming process undertaken using spreadsheets. Recent advances

    in the Whittle software have enabled multiple mines to be optimised andscheduled simultaneously, so that the mining sequence that maximises theNPV (net present value) for the entire set of mines as a whole can be

    determined. This case study presents the results of the collaborationbetween Gemcom Africa (Pty) Ltd and Geita Gold Mining Limited todevelop and apply the Whittle Multi-Mine module at Geita Gold Mine. It

    shows how improvements to the NPV of the life of mine schedule wereachieved by using Whittle Multi-Mine as a tool to help guide the

    preferred order of mining. It highlights the contributions from each of themines to the overall cash flow of the project and investigates the effect of

    time on the NPVs from each mine. The cost of deferring production fromcertain mines has become plainly evident, whilst for others there is littleimpact. Furthermore, Whittle Multi-Mine has identified areas requiringmore focus in terms of the life of mine plan.

    INTRODUCTION

    Geita Gold Mine is situated in northwest Tanzania,approximately 25 km from the southern shores of Lake Victoria.Historical mining in the area has taken place for many years,with the last major operation being the Geita Underground Mine,which operated from the 1930s through to the 1960s andproduced almost 1 Moz of gold. Ongoing small scale miningcontinues to this day.

    The modern Geita mine has been operating since mid-1999,

    with processing of ore commencing in mid-2000. To date, 48Mbcm of material has been mined from three open pits; 14 Mt ofore, grading 3.8 g/t, has been processed and 1.5 Moz recovered.The current Life of Mine Plan indicates a mine life in excess often years and entails the mining of ten individual pits, several ofwhich are multi-stage. Total mining is expected to exceed 320Mbcm, producing more than 80 Mt of high-grade ore andyielding more than 10 Moz of recovered gold.

    The open pit mines are operated with conventional techniquesusing excavators and trucks on flitches up to 3.5 m high. Mostmaterial requires blasting, ranging from paddock blasting insoft laterites and oxides, to hard rock blasting in sulfides.

    Pit optimisation at Geita has been an ongoing process,predominantly undertaken using the NPV Scheduler software,however; from early 2003 Whittle software has been used inparallel. Although techniques to evaluate multiple orebodies haveexisted for some time (Tulp, 1997), each open pit has beenoptimised and scheduled as a separate entity rather thanconsideration given to whole of mine optimisation andscheduling. Estimates of the mill throughput likely to be requiredfrom each pit were used to guide the pit life and net present value(NPV) calculation. Since the ore delivery rate required wasgenerally not known until the whole mine schedule was finalisedusing all the pits, this was obviously a flawed process.

    Once the optimal pit for each mine was decided, pit designswere undertaken, reserves calculated and the entire data setexported to a spreadsheet for manual scheduling. Variousguidelines and comparisons between the pits and stages wereused to assist with the manual scheduling process, such as stripratio, profit per tonne milled, cash cost per ounce, profit perounce and break-even time. This introduced another flaw in theprocess, where the optimal extraction sequence was notnecessarily followed during the manual scheduling process.

    It became apparent that this trial and error scheduling methodwas time-consuming and limited the number of alternate life ofmine scenarios that could be evaluated. A need for a technique tooptimise the extraction sequence in this multiple mine scenariowas identified. Such a tool was available as part of the Whittle

    suite of mine planning software, but was still in its infant stages,requiring rigorous testing on a real life scenario. This paperdescribes Whittle Multi-Mine and its application at Geita, butfirst briefly reviews the traditional and widely applied MOBStechnique.

    MOBS

    A technique known as MOBS (Multiple Ore Body Systems,Tulp, 1997) has existed for some time now and has been widelyapplied in situations where multiple orebody deposits exist inproximity. In short, the technique involves agglomerating blockmodels representing each of these deposits into one super model(Figure 1), and optimising and scheduling using Whittlesoftware. The limitations of this method are described following.

    To enable the identification of material selected for mining byWhittle from the different deposits, it was necessary to assignunique rock codes that were reflective of the different depositareas. Furthermore, the rock codes used for Whittle also neededto capture the actual rock type, so that different mining andprocessing costs could be defined if necessary. This required theassignation of many rock codes and sometimes resulted in theloss of geological definition due to the restriction in the numberof codes that could be handled by Whittle.

    Once the optimisations had been completed and the pit shellsgenerated, it was necessary to cut up the super model results fileto separate the individual mines, using the polygon intersection

    Orebody Modelling and Strategic Mine Planning Perth, WA, 22 - 24 November 2004 267

    1. MAusIMM, C/- Gemcom Africa (Pty) Ltd, PO Box 411689,Craighall 2024 Gauteng, Republic of South Africa.

    2. MAusIMM, C/- MRM Mining Services, PO Box 3193, HalfwayHouse 1685 Gauteng, Republic of South Africa.

    3. 66 Rathmullen Quadrant, Doncaster Vic 3108.

    NORTH PIT

    WEST PIT

    MAIN PIT

    SATELLITE 1 SATELLITE 2

    FIG 1 - Example super modelconstructed by merging

    MOBS in Whittle.

  • 8/22/2019 Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

    2/6

    functionality in Whittle, so that the results could be exportedfrom Whittle back to a general mining package (GMP). This wasbecause the original coordinates of the individual deposits werelost when they were combined into the super model.

    Issues arose when scheduling MOBS, since it was not possibleto control the order that the deposits were mined in withoutcreating complex pit list files with GMPs or by writing scriptswith programming packages. Furthermore, it was necessary to

    ensure that the top surface of all of the models lay on the sameWhittle bench level in the combined super model, requiring theuser to offset each individual block model so as to create aregular surface over the entire model. This meant that whensimulating the mining of a bench in Whittle, the bench wasmined from all of the mines in the super model. It was also notpossible to have different cut-backs in each mine, nor was itpossible to have different final pits per mine. This reduced theeffectiveness of the scheduling and did not allow areas of highervalue to be deliberately targeted.

    For more advanced scheduling using the Whittle Milawascheduling algorithm, it was necessary to stack groups of pitshells, representing the nested pits derived for each mine, forMilawa to work effectively (Figure 2). This was difficult to set upand comparatively inflexible when evaluating many alternate

    mining sequences.Whilst the technique described above generated results that

    added value to mining operations; it was tedious and much timewas spent on manipulating models and data files, thus limitingthe amount of time that could be spent on actually evaluatingdifferent scheduling sequences and the consequent impact onNPV.

    MULTI-MINE

    Whittle Multi-Mine provides a much more sophisticated andflexible means of optimising and scheduling in a multiple minesituation, as was proven by its successful application at Geita

    Gold Mine. The different techniques applied at Geita aredescribed following, using examples (Joukoff and Purdey, 2004)

    to illustrate the results.

    With Whittle Multi-Mine it is no longer necessary to use rockcodes to identify material from different deposits. It is nowpossible for Whittle model files to carry a mine name, so the

    issue of running out of rock codes is no longer a problem. Thisallows greater geological detail to be modelled, leading toincreased flexibility and detail when modelling costs, recoveries

    and slopes in Whittle, if desired. Furthermore, because eachmodel can be associated with a mine name, it is possible to view

    and export results for individual mines. This reduces the amountof time required to be spent on data manipulation and providesmore time to deal with strategic issues.

    It is possible to optimise all the mines under considerationeither simultaneously or individually, because the Whittle modelfiles carry a mine name. The advantage of optimising them

    together is that the impact of each mine on the combined cashflows of all the mines can be examined and reported.

    Scheduling with Multi-Mine is now also much more

    sophisticated than the MOBS technique previously applied. It ispossible to vary the mining rates in different mines and also tocontrol when mining can occur in a particular mine. This

    functionality proved particularly useful at Geita because some ofthe mines were remote from the processing plant and ore

    production from these mines was limited by the long distancehaulage capacity (Figure 3). Also, due to Geitas environmentalcommitment to backfilling completed pits to minimise

    disturbance caused by the construction of waste dumps, somemines were not able to commence until adjacent mines werecompleted. Furthermore, either of the Fixed Lead or Milawa

    scheduling algorithms can be applied as described following.

    268 Perth, WA, 22 - 24 November 2004 Orebody Modelling and Strategic Mine Planning

    T JOUKOFF, D PURDEY and C WHARTON

    FIG 2 - Stacked pit shells to enable Milawa to operate

    independently on each mine before the development of Whittle

    Multi-Mine.

    Roberts

    Ridge 8 Star/Comet

    Geita Hill Pits

    Nyankanga

    Chipaka

    MatandaniKukuluma

    Lone Cone Pits

    Legend

    Haul Road

    Pit

    Plant

    Schematic of Geita Gold Mine(Not to scale)

    ~5 km

    ~20 km

    ~19 km

    FIG 3 - Schematic of Geita Gold Mine.

  • 8/22/2019 Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

    3/6

    Fixed lead

    Fixed lead scheduling can operate with or without precedencecontrols. By establishing mining precedence rules, different

    orders of mining the individual mines can be simulated, makingit possible to investigate which order maximises the NPV to thecompany. This technique is particularly applicable in situations

    where only one mine will operate at a time, such as when themines are very large and where ore control issues can be handledsufficiently by manipulating the mining sequence within each

    mine, without the need to blend material from different mines.Each mine may have its own process plant and associatedinfrastructure but logistically, mining equipment may need to

    move from one mine to another. The order of mines to whichequipment moves can be optimised using this functionality.Alternately, when no particular precedence is required and

    mining can occur simultaneously in all mines following the samebench lead constraints, fixed lead scheduling can also be applied.

    These two alternate concepts are illustrated in Figure 4.

    Fixed lead scheduling was tried at Geita but with limitedeffectiveness because the site wanted to be able to mine from

    many pits simultaneously, rather than mine them sequentially.Although this was possible as described previously, it was not

    practical in Geitas case because several of the mines werealready in production and operating on different bench levels.Furthermore, within the constraints of the existing cut-backdesigns at Geita, using fixed lead scheduling did not provide an

    optimal mill feed schedule. Geita needed to be able to drawmaterial from multiple sources to feed the mill, to meet theappropriate oxide/sulfide blend requirements and also to make

    better use of the available mill capacity. Greater flexibility wasrequired, and to overcome these issues it was necessary to apply

    the Milawa algorithm.

    Milawa

    The majority of the Geita scheduling work in Multi-Mine wasundertaken using the Milawa scheduling algorithm. This was

    because Milawa allowed material to be mined from differentmines simultaneously, applying different lead and lag constraintsto the different mines (as opposed to fixed lead scheduling,which uses the same lead constraint for each mine). There was arequirement at Geita to limit the maximum highwall heightbetween cut-backs to 150 m, for geotechnical reasons. Themaximum vertical advance in each mine was also restricted toeither 50 m or 100 m per year, depending on the size of the mine.For this reason it was necessary to define different constraints fordifferent mines and this was easily achieved with Multi-Mine.

    It would be prudent at this stage to briefly explain thedifferences between the various Milawa scheduling algorithms.In NPV mode, Milawa will seek to maximise the NPV of the

    schedule, taking into consideration the number of benches,

    cut-backs and time periods in the life of the mine (Wharton,

    2000). Milawa NPV schedules generally mine just enough wasteto uncover the ore required to fill the mill and tend to defer wastestripping as much as possible. Logically, this will lead to

    increased NPVs. However, this waste deferral may result ininsufficient ore availability at some time in the schedule, but onlyif the cut-backs have not been selected appropriately or if the

    mining capacity is not well matched to the selected cut-backs.

    The Milawa algorithm in balanced mode provides a solution to

    this problem by producing a schedule that completely utilises allof the available mining and milling capacity where possible. Thegeneral effect of such a schedule is to mine more waste than is

    needed to uncover the ore necessary to feed the mill, hencebringing costs forward and resulting in a reduced NPV. However,both the mill feed schedule and the total mining schedule will be

    well balanced. A diagrammatic sketch of a Milawa miningsequence is included in Figure 5.

    APPLICATION AT GEITA GOLD MINE, TANZANIA

    Geita Gold Mining Limited provided a data set representing nineof the mines planned as at November 2003 (Nyankanga, Lone

    Cone, Geita Hill, Kukuluma, Matandani, Chipaka, Ridge 8,Star/Comet and Roberts). Each model was exported from a GMPwith pre-defined rock types that allowed unique costs and

    process recoveries to be assigned to each rock type. Although itis possible to model costs in Multi-Mine using a Mine variable,a cost model reflecting the different long distance haulage costs,

    defined for different rock types, already existed. As well as this,the existing cut-back positions were exported as pit list models,allowing the cut-backs within each mine to be differentiated

    during subsequent analysis. These pit lists were agglomerated inWhittle to create a results file suitable for use with theMulti-Mine scheduling tools. Some of the required operational

    constraints have already been described previously in this paper.

    Before undertaking any further scheduling in Whittle, a

    baseline schedule was developed with Multi-Mine that mimickedthe existing Life of Mine (LoM) Plan as much as possible. This

    was so that subsequent NPV calculations for alternate miningsequences would be comparable. An iterative process was usedin defining this baseline schedule, using modifications to themin/max lead and max benches constraints to force Multi-Mine

    to mine in a similar sequence and with similar quantities asdefined in the LoM Plan. Concurrent with this work inMulti-Mine was the recalculation of the LoM Plan NPV because

    this included the effects of many cash outflows that were notapplicable in pit optimisation.

    Once the Multi-Mine baseline schedule was constructed, theconstraints were selectively relaxed to allow Multi-Mine to beginto optimise the schedule. Alternate orders of mining were tested

    by simply adjusting the preferred order of mining and the minestart and stop times, and the resultant NPV, ore delivery schedule

    and total mining schedule evaluated.

    Orebody Modelling and Strategic Mine Planning Perth, WA, 22 - 24 November 2004 269

    DEVELOPMENT AND APPLICATION OF WHITTLE MULTI-MINE AT GEITA GOLD MINE, TANZANIA

    M1 M2 M3M1 M2 M3 M1 M2 M3M1 M2 M3

    A. B.

    FIG 4 - Diagrammatic representation of two different fixed lead scheduling sequences in Whittle Multi-Mine. A. Mining precedence applies

    and equipment moves from one mine to another on completion of each mine (Wharton, 2000). B. No mining precedence applies and all

    mining occurs simultaneously in all mines, following specified bench lag constraints (Wharton, 2000).

  • 8/22/2019 Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

    4/6

    In total, twenty-four different LoM scheduling scenarios forGeita were considered using the Milawa algorithm in

    conjunction with Multi-Mine. Comparison of the NPV of each ofthese schedules with the baseline schedule showed that the NPVsranged from 87 per cent to 103 per cent of the baseline NPV.

    Whilst a three per cent improvement in NPV may seem small, inGeitas case it represented an increase in NPV in excess of 1500times the cost of undertaking the Multi-Mine work. An ore

    schedule representative of the results generated with Multi-Mineis displayed in Figure 6.

    The most significant difference between the Whittle

    Multi-Mine results and the existing site LoM Plan was that theMilawa algorithm preferred to mine Star/Comet as early as

    possible, rather than later in the project life as had beenpreviously scheduled. This gave some indication as to thesignificance of the Star/Comet mine to the overall project NPV.

    When run unconstrained, Multi-Mine also preferred to mineMatandani in early years, but this was not a favoured option asthe waste from Matandani was planned to be backfilled into the

    Kukuluma mine.

    Investigation of the contribution to NPV from each mine foreach scheduling scenario helped to determine which mines the

    overall NPV was most sensitive to. Table 1 contains arepresentative set of results showing these cash flow

    contributions for various scenarios. It is clear that for some of the

    mines changes to the order of mining had little or no effect ontheir contribution to total NPV, whilst for others the change in

    contribution to NPV was considerable.

    The effect of delaying production from any mine can be seen.The cost of deferring Nyankanga is very evident; the NPV

    contribution being as much as 67 per cent (Scenario 14) or as

    little as 46 per cent (Scenario 3). This represents a 21 per centimprovement in cash flow contribution from Nyankanga forScenario 14 compared with Scenario 3. In fact, in Scenario 3 the

    NPV from Nyankanga approaches that of worst case mining.As a further example consider Chipaka mine; if this is mined last(Scenario 14) the NPV contribution erodes to just 0.5 per cent,

    but if it is mined first (Scenario 17), the NPV contribution can beas much as two per cent. However, when considering the NPV of

    all of the mines concurrently, delaying Chipaka gives the projecta better overall NPV. This clearly demonstrated how the order ofmining can have a serious impact on the value of the project.

    It was concluded from all of the scenarios that the NPV wasrelatively insensitive to changes in the order of mining from theChipaka, Kukuluma and Ridge 8 mines. This suggested that it

    was not worthwhile to further optimise the timing of these mines.Conversely, there was substantial gain to be made by optimisingthe mining sequence from Nyankanga, Geita Hill, Matandani and

    Star/Comet. For this reason, the order of mining from thesemines was the focus for the remainder of the scenarios andyielded higher value schedules.

    Examination of the bench schedules produced by WhittleMulti-Mine helped to understand how much material was mined

    from each bench, each cut-back and each mine in each periodand hence made it possible to determine whether Multi-Minewas adhering to the required operational constraints. The

    resultant schedules were both safe and practical. Furthermore, bymaking comparisons between the benches mined in differentscheduling scenarios it was possible to understand where the

    material was being mined from, and the subsequent contributionof that material to the overall value of the schedule. An examplebench schedule is given in Table 2.

    CONCLUSIONS

    This paper has reviewed the techniques available in Whittle tooptimise and schedule multiple orebody models and multiplemines. The application of Whittle Multi-Mine at Geita Gold

    Mine, Tanzania, has demonstrated how improvements to theNPV of the life of mine schedule were achieved, usingMulti-Mine to help optimise the mining sequence. The Milawaalgorithm in both NPV and balanced mode was able to guide theorder of mining benches from the various cut-backs of thevarious pits, within the operational constraints at Geita.

    270 Perth, WA, 22 - 24 November 2004 Orebody Modelling and Strategic Mine Planning

    T JOUKOFF, D PURDEY and C WHARTON

    M1 M2 M3M1 M2 M3

    FIG 5 - Diagrammatic sketch of a Milawa mining sequence inWhittle Multi-Mine (Wharton, 2000).

    MILL FEED SCHEDULE

    TONNAGE

    GRADE

    YEAR

    FIG 6 - Representative ore schedule, Geita Gold Mine case study. Different shades represent different mines.

  • 8/22/2019 Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

    5/6

    Many alternate scheduling sequences were very quicklyinvestigated using Whittle Multi-Mine. This process identifiedwhich mines demonstrated greater sensitivity to the order inwhich they were extracted and subsequently stressed the effect oftime on the cash flow contribution of these mines to the overallproject NPV. It also assisted in highlighting a potential mismatchbetween the required material movement and the availablemining capacity. If the mining capacity is well matched to theselected cut-backs then it will be possible to achieve a balancedschedule together with an improved NPV.

    ACKNOWLEDGEMENTS

    This paper describes work undertaken by co-author DavidPurdey whilst employed as Chief Mining Engineer Geita GoldMining Limited and is presented with Geita Gold MiningLimiteds permission. The authors would like to thank GeitaGold Mining Limiteds management for their permission topresent this paper and also thank the members of the Mining

    Department at Geita who contributed to the preparation of thedata used in the Multi-Mine analyses.

    The opinions expressed in this paper are not necessarily thoseof Geita Gold Mining Limited.

    REFERENCES

    Joukoff, T and Purdey, D P, 2004.http://www.whittle.ca/whittle-multimine.asp Improved life of minescheduling with Gemcom Whittle Multi-Mine at Geita Gold Mine,Tanzania (Gemcom Software International Inc: Vancouver).

    Tulp, T, 1997. Multiple Ore Body Systems (MOBS), in ProceedingsOptimising with Whittle, Perth, pp 149-163 (Whittle ProgrammingPty Ltd: Melbourne).

    Wharton, C, 2000. Add value to your mine through improved long termscheduling, in Proceedings Whittle North American Strategic MinePlanning Conference, Breckenridge, Colorado.

    Wharton, C, 2003. Multi-pit analysis and advanced pit scheduling,Development notes (unpublished), Melbourne.

    Orebody Modelling and Strategic Mine Planning Perth, WA, 22 - 24 November 2004 271

    DEVELOPMENT AND APPLICATION OF WHITTLE MULTI-MINE AT GEITA GOLD MINE, TANZANIA

    Pit Bench Total Year 1 Year 2 Year 3

    Tonnes Ore t Waste t Ore t Waste t Ore t Waste tKukuluma 69 1496 - - 341 1155 - -

    Kukuluma 68 1131 - - 306 825 - -

    Kukuluma 67 772 - - 244 527 - -

    Kukuluma 66 453 - - 194 259 - -

    Kukuluma 65 20 - - 12 9 - -

    Kukuluma 64 142 - - - - 105 37

    Kukuluma 63 56 - - - - 44 12

    Kukuluma 62 11 - - - - 9 2

    Subtotal

    Lone Cone 66 1266 161 1105 - - - -

    Lone Cone 65 1176 201 975 - - - -

    Lone Cone 64 1015 154 861 - - - -

    Lone Cone 63 901 111 789 - - - -

    Lone Cone 62 788 - - 73 715 - -

    Lone Cone 61 653 - - 44 609 - -

    Lone Cone 60 532 - - 37 495 - -

    Lone Cone 59 401 - - 43 358 - -

    Subtotal

    TABLE 2

    Example extract from bench schedule generated using Whittle Multi-Mine.

    Matan' Chipaka Geita Hill Kuk' Lone Cone Ridge 8 Roberts Star Comet Nyank' Total

    Scenario 2 5% 1% 17% 4% 3% 1% 1% 4% 63% 100%

    Scenario 3 5% 2% 28% 5% 4% 1% 3% 6% 46% 100%

    Scenario 7 5% 1% 17% 4% 3% 1% 1% 9% 57% 100%

    Scenario 14 3% 0% 16% 4% 2% 1% 2% 6% 67% 100%Scenario 15 3% 1% 18% 4% 3% 1% 1% 8% 60% 100%

    Scenario 16 4% 1% 17% 4% 3% 1% 2% 8% 60% 100%

    Scenario 17 3% 2% 16% 4% 3% 1% 1% 6% 65% 100%

    Scenario 19 4% 1% 18% 4% 4% 1% 2% 8% 59% 100%

    Scenario 21 5% 1% 22% 5% 4% 1% 2% 7% 55% 100%

    Scenario 23 5% 1% 20% 5% 1% 1% 3% 9% 56% 100%

    TABLE 1

    NPV contributions by pit by scenario, Geita Gold Mine case study.

  • 8/22/2019 Application_of_Whittle_Multi_Mine_at_Geita_Gold_Mine_T.Joukoff_et_al.pdf

    6/6

    272 Perth, WA, 22 - 24 November 2004 Orebody Modelling and Strategic Mine Planning