05708 mill

8/13/2019 05708 mill

1/16

III-11

SAG MILL CIRCUIT OPTIMISATION AT ERNEST HENRY MINING

S. Strohmayr1, W. Valery

2Jr.

1Ernest Henry Concentrator Operations Superintendent

2Julius Kruttschnitt Mineral Research Centre Project Leader

ABSTRACT

The Ernest Henry Mine is situated approximately 35km north-east ofCloncurry in the Mount Isa Cloncurry mineral district of north-westQueensland. The concentrator is a single line plant using contemporaryequipment with a nominal throughput rate of 1200 tonnes per hour. Anaverage of 350,000 tonnes of concentrate is produced each yearcontaining 100,000 tonnes of copper metal and 120,000 troy ounces ofgold.

The milling circuit was commissioned in August 1997 and consists of a10.4m x 5.1m SAG mill powered by two 5.5 MW drives, in closed circuit

with a 3.7m x 7.3m vibrating screen. Screen oversize is recirculatedback to the SAG feed with conveyors. The screen undersize combineswith the ball mill discharge in a common sump and is pumped to 840mmprimary hydrocyclones. The hydrocyclone overflow advances to rougherflotation while the hydrocyclone underflow forms the ball mill feed. The6.4m x 8.1m ball mill has 5.5MW of installed power.

An optimisation project was conducted with the Julius KruttschnittMineral Research Centre (JKMRC) to optimise SAG mill circuitperformance. Alternative circuit operating conditions were investigatedand SAG mills lifter/liner design was modified. JKSAGCharge wasinstalled to control the SAG mill charge. The project also included an

investigation of the effect of run of mine fragmentation in SAG millthroughput and based on the simulations, alternative blast designs were

III-12

suggested to maximise SAG mill throughput. This paper describes thedata and results of the optimisation project.

GENERAL

Ernest Henry Mining Pty Ltd (EHM) is an incorporated joint venturebetween MIM Holdings Limited (51%) and Investco Ltd. (49%). The

Ernest Henry Mine ore reserve is 123 million tonnes averaging 1.1 wt%Cu and 0.55 g/t Au. The mineralised zone can be broadly subdividedinto two main zones; the supergene and the primary. Primary ore,which comprises approximately 85% of the orebody, is mineralogicallysimple while supergene ore mineralogy is complex.

The life of mine strip ratio is 4.8:1, however the ratio varies from 2.5:1 inthe first year of pit operations up to 5.6:1 in Year 8, to less than 1:1 inthe last 2 years of production. The final pit design has a surfacediameter of 1300m and a depth of 570m. The scale of the project andorientation of the deposit lends itself to open cut extraction using largeequipment.

The concentrator, commissioned in August 1997, is a single line plantusing contemporary equipment with a nominal throughput rate of 1200tonnes per hour. An average of 350,000 tonnes of concentrate will beproduced per year containing 100,000 tonnes of copper metal and120,000 troy ounces of gold.

In the financial year just completed the concentrator milled 10.3 milliontonnes at 1300 tonnes per hour to produce 105,000 tonnes of coppermetal and 118,000 troy ounces of gold.

The majority of the copper concentrate is trucked to Mount Isa for

smelting.

MINERALISATION

The following geological description is largely from a paper published byRyan (1998).

Mineralisation at Ernest Henry is developed in a south-east plungingbody of altered and variably brecciated felsic volcanic rock. Thecombined thickness of the mineralised sequence is approximately250m, width averages 300m and the down dip length is approximately1000m. The down dip limit remains untested at this time.

Mineralisation can be divided into two main zones; the supergene andthe primary. The supergene zone comprises material that has been

8/13/2019 05708 mill

2/16

III-13

modified by weathering processes and therefore lies above the Base ofPartial Oxidation (BOPO). The primary zone lies below the BOPO andshows no effect of weathering. Supergene ore makes up 15% of the orebody and primary ore makes up the remaining 85%.

Primary Zone

The primary ore mineralogy is quite simple. The ore assemblage isdominated by chalcopyrite within a magnetite-carbonate gangue, thereare no other oxides or sulphides of economic importance. Coppergrades show a bi-modal distribution, with the

8/13/2019 05708 mill

3/16

III-15

39500mN

39300mN

39100mN

38900mN

38700mN

68800mE

69000mE

69200mE

69400mE

69600mE

Ultimate Pit

Boundary

Surface

Projections

of Orebody

Upper Lens

Lower Lens

Eastern Lens

ML2671

Figure 1 Location of Drill Holes used in Testwork

1. The initial metallurgical characterisation of the ore indicated that it fell

within the very substantial body of knowledge for porphyry copper ores,2. A significant amount of time would be added to the developmentschedule, and3. The pilot plant testing would provide significant amounts of data aboutthe parcel of material tested, but not provide data for the orebody as awhole. The ability to obtain a representative sample was considereddifficult if not impossible.

Therefore the testwork effort was directed towards gathering as muchdata about the orebody through the use of laboratory tests.

Grinding Testwork

The following organisations performed grinding testwork:

III-16

Julius Kruttschnitt Mineral Research Centre (JKMRC) A.R. McPherson Consultants (ARM) AMDEL Orway Mineral Consultants (OMC) Met Engineers Ltd (Canada)

In addition, Fluor Daniel undertook in-house evaluations of the grindingcircuit testwork results.

Samples from five 85mm diamond drill holes were used in thecomminution tests at JKMRC, AMDEL and ARM. These results alongwith Bond Work Index tests carried out by Met Engineers Ltd as part ofthe flotation testwork, were the basis of the initial mill sizing by OMC,

ARM, JK Tech and Fluor Daniel Wright (FDW).

Point Load Index tests (PLI) were carried out on the drill core from thegeological drilling program to provide data on the variance in grindingcharacteristics over the orebody.

Primary ore is fairly homogeneous, and its response to comminutionwas relatively uniform on the samples tested. Since primary oreconstitutes approximately 85% of the total ore body the grinding circuitdesign is based on the treatment of this ore. The supergene ore zone isconsiderably more variable in composition and physical characteristics,which posed a problem for the grinding circuit design since thesupergene material is the major ore type for the first three years ofproduction. Circuit flexibility was incorporated as a fundamental designcriterion.

COMMINUTION TEST RESULTS

The following is a listing and interpretation of the results from thelaboratory grinding testwork.

Bond Work Index Testing

Table 1 on the next page summarises the results from the Bond WorkIndex testing. The results showed that the ore was relatively consistentin grindability, particularly in the ball milling range. The variability in therod mill work indexes was much more significant and indicated thatsome of the primary ore and a greater amount of the supergene ore isless competent than the average. Hence a reasonably high ball load in

the SAG mill would be required to achieve the design throughput rateswith a satisfactory product size.

8/13/2019 05708 mill

4/16

III-17

Table 1: Bond Work Index Test Results

Work Index (kWh/t) Primary Ore Supergene OreRod Mill Average 11.4 10.1

Maximum 15.2 13.6Minimum 8.9 7.6

Standard Deviation 2.1 1.7

Ball Mill Average 11.7 10.7Maximum 15.1 12.4Minimum 9.8 8.4

Standard Deviation 1.8 1.2

Ore Specific Gravity

Specific gravity determinations are shown in Table 2 below.

Table 2: Ore Specific Gravity

Primary Ore SG Supergene Ore SG

Maximum 3.73 3.38Minimum 3.05 2.96

Average 3.42 3.25Mill Charge 3.12 3.29

The average feed densities were used for mill sizing, with allowances foroperation over a range of densities.

Unconfined Compressive Strength.

The unconfined compressive (UCS) values obtained are shown in Table

3.

Table 3: Unconfined Compressive Strength

Primary Ore (MPa) Supergene Ore (MPa)Maximum 161 110Minimum 79 20

Average 116 55

Interpretation of these results indicated that the supergene ore was lowin competency and would require a high SAG ball to achieve throughput

rates, while the primary ore was quite competent and would producegood SAG mill media.

III-18

JKMRC Drop Weight Tests

JKMRC drop weight tests were conducted and the results are presentedin Table 4.

Table 4: JK MRC Comminution Parameters

Sample ParametersA b A*b ta

Primary Ore 52 0.78 41 0.4777 0.60 46 0.33

Supergene Ore 78 0.53 41 0.7855 1.98 109 1.19

Compared to the average SAG/FAG appearance function data in the JKTech database, primary ore and the harder supergene ore is lesssusceptible to impact breakage (harder) with a lower-than-averagetendency to produce fines by abrasion.

However, there are many ores, which are more resistant to both impactand abrasion which are being successfully processed by AG/SAGmilling eg. MIM copper ore has b values as low as 0.81 (A=50 at

A*b=40) with lower ta values at 0.28 0.33.

The softer supergene ore is substantially softer than average databaseSAG ores tending to break down more by impact with a substantiallygreater tendency to abrade to create fine particles.

The results listed above were interpreted as follows;

Supergene ore has low competency with little chance of successfulautogenous milling (or grinding), while primary ore is generallytough and competent, particularly in the 50 to 75 mm size rangeswhich makes excellent grinding media. The primary ore wouldprobably produce sufficient media for fully autogenous grinding.

Both supergene and primary ores have below average compressivestrengths and above average grindability.

It would be possible to treat primary ore autogenously, but safer touse a Semi Autogenous Grinding (SAG) mill ball charge of 8% v/vfollowed by a ball mill.

8/13/2019 05708 mill

5/16

III-19

Grinding Flowsheet Selection

The philosophy of a single grinding line using some of the largestcommercially available equipment was adopted in order to keep theflowsheet and plant layout simple with the objective of minimising costs.

ARM, OMC and FDW have all recommended a flowsheet consisting of

SAG milling in closed circuit with a vibrating screen followed by a ball millin closed circuit with hydrocyclones. There was strong agreementbetween all consultants that this would represent the lowest risk option,considering the nature and scale of samples tested.

An alternative circuit, consisting of a fully autogenous mill in closedcircuit with a recycle crusher followed by a ball mill in closed circuit withhydrocyclones (ABC circuit), was also considered to be suitable forprimary ore. OMC estimated the circuit would consume about 20% morepower than the SAG/ball circuit, which approximated the savingsobtained by the elimination of grinding media. However, the ABC circuitwould not be suitable for treating supergene ore due to its low

competency and was estimated to have a higher capital cost.

An extensive review of the alternative circuits revealed that the differencein capital cost between the various options was insignificant. To reducethe uncertainty associated with the risk it was decided to proceed with avery conservative selection. A SAG/ball mill circuit using the largestcommercially available ball mill in conjunction with a proven SAG milldesign was selected as the preferred option. This circuit would besuitable for treating both primary and supergene ores.

Mill Sizing for Primary Ore

Three independent consulting groups made mill size selection for thetreatment of 9 Mt/a primary ore. Table 5 on the next page summarisesthe results.

There was reasonable agreement between the selections. However, thelow total power proposed by ARM indicates that a SAG/ball combinationwould be as power efficient as a rod/ball mill circuit. This is unlikely to bethe case for a SAG mill in closed circuit with a vibrating screen.

A single SAG mill/ball mill circuit based on the FDW mill sizing wasselected. This sizing used a ball mill work index of 10.0 kWh/t based onthe 1992 testwork. The further grinding testwork conducted gave an

average ball mill work index of 11.7 kWh/t. The mill sizes weresubsequently modified to reflect the increased power.

III-20

Table 5: Comparison of Consultants Mill Sizing Recommendations

OMC ARM FDWSAG Mill No. of Units 1 1 1

Diameter m 9.75 10.37 10.37

Length m 4.27 4.87 4.42Power Draw kW 6582 7986 8842

Motor Power kW 8250 8500 10000

Ball Charge %v/v 8 8 10Work Index kWh/t 29.32 12.7 20.46

Transfer Size k80 m 1970 300 610

Ball Mill No. of Units 2 1 1

Diameter m 5.33 5.18 5.49Length m 7.62 9.15 8.38

Power Draw kW 6650 3013 4700Work Index kWh/t 12.4 12.0 10.0

Total Power Draw kW 13232 10817 13542

The final mill sizes selected were undertaken with the above in mind

and considerations of commonality of parts between the mills. The millsselected were:Table 6: Final Mill Selection

SAG Mill the worlds largest shell supported SAG Mill. ( Tew, 1999)No. of Units 1Diameter m 10.4Length m 5.1Motor Power kW 11,000Ball Charge % v/v 8Work Index kWh/t 10.7Transfer Size T80 m 610

Ball MillNo. of Units 1Diameter m 6.10Length m 8.4Power Draw kW 5,500Work Index kWh/t 11.7

Total Power Installed kW 16,500

Grinding Simulation

JK Tech were engaged to run simulations on this selection andconfirmed that a suitable product could be obtained from a circuitcomprising the selected SAG mill in closed circuit with a single deck

8/13/2019 05708 mill

6/16

III-21

vibrating screen, combined with a ball mill in closed circuit withhydrocyclones.

The simulations highlighted that classification, using hydrocyclones, ofthe ball mill circulating load to produce a coarse product would bedifficulty due to the relatively high specific gravity of the hydrocyclonefeed solids. The hydrocyclones would have to be operated very

inefficiently to achieve the desired result. This would result in highclosed circulating loads of dense minerals (magnetite and native copper),overgrinding and inefficient use of ball mill power.

Supergene Ore Treatment

The variable supergene ore performance was assessed andconservative estimates showed that it would be milled at a rate 10 to15% greater than the primary ore although the conditions in the SAG millwould have to be altered. It was envisaged that ball charge, mill speedand recirculation of the hydrocyclone underflow could be used as controlvariables. The SAG mill was installed with a fixed speed motor, however

the possibility to retrofit a variable speed drive with limited speedadjustment, if it is found to be necessary, was allowed.

The final process flowsheet is shown in Figure 2 below.

ERNEST HENRY MINING PROCESS FLOWSHEET

mine

Gyratory

Crusher Stockpile

200,000 tonne capacity

SAG Mill

Svedala Krupp 11 MW drive

60" x 89" 10.4m x 5.1 m

Screen

7mmaperture

Rougher

9x127m3Wemco Smart Cells

Krebs

Cleaner 1Krebs 8x 840mm

8x50m 3 OK50's 6x250mm

Cleaner 2

8x16m3OK16

Svedala

1 MV drive

Cleaner 3

5x16m3OK16 Verti Mill Screen Ball Mill

Krupp 5.5 MW drive

6 .1mx8 .4m

Native Copper Stockpile

Tailings Thickener Concentrate Thickener Pressure

Eim co 55 m tr ac tio nd riv e E im co 25 m Filter

Concentrate Stockpile

Tailings Dam Larox144m2

Figure 2: Ernest Henry Flowsheet

III-22

COMMISSIONING

The plant commissioning was reasonably uneventful. This wasattributed to a number of factors;

The approach to commissioning was very structured. The plant wasdivided into small modules where detailed programs of work were

undertaken by dedicated commissioning crews. As the modules werecomplete they combined with others to form larger units of the plant untila section, for example grinding, was signed off ready to run.

Employment of experienced people at all levels of operation well beforestartup. Where possible operations personnel were involved incommissioning.

Figure 3 shows the throughput and plant availability over the first 19months of operation. Throughput and plant availability targets were metand exceeded within the first year of operation.

Ernest Henry Mining - The First 19 Months

Plant Availability - Ore Milled

0

100,000

200,000

300,000

400,000

500,000

600,000

700,000

800,000

900,000

Aug-97

Sep-97

Oct-97

Nov-97

Dec-97

Jan-98

Feb-

98

Mar

-98

Apr-9

8

May

-98

Jun-98

Jul-9

8

Aug-98

Sep-98

Oct-9

8

Nov-98

Dec-98

Jan-99

Feb-

99

MonthlyThroughput(dmt)

0

10

20

30

40

50

60

70

80

90

100

Availability(%)

Milled Tonnes - Actual Plant Avai labil i ty (%) Feasibi li ty Forcast Throughput

Design Plant Availability 90.5%

Figure 3: Feasibility Predictions vs Actual Production Results

FIRST YEAR OF OPERATION

Ore Variability

Supergene ore was the sole source of feed to the concentrator for thefirst 6 months of operation. The short-term variability (1 2 hrs) of the

8/13/2019 05708 mill

7/16

III-23

ore in terms of throughput, head grade and metallurgical response farexceeded the predictions from the feasibility testwork.

Throughput typically varied by up to 800 tph within a single 12-hour shift.SAG milling is particularly sensitive to changes in feed characteristics,thus the variation seen in the mill feed characteristics had significantdetrimental affects on steady SAG mill operation. Figure 4 at the top of

the next page shows a typical 12 hour SAG trend.

Additionally the SAG mill ball load could not be set up optimally fortreating the harder ores, because the ore could just as readily change toeasy milling material, which when treated with a high ball charge emptiedthe mill of rock charge, resulting in damage to the shell liners. The rapidchange in ore hardness resulted in the significant shell liner failuresexperienced during the first 6 months of operation.

Operational techniques and process modifications incorporated toprevent the mill emptying on softer ore included:

Running the mill at high mill wt% solids > 88 wt% solids. Operating with a ball charge volume of 3 4 % v/v. Operating the SAG mill apron feeders in a manner to produce the

coarsest feed size. (qualitative). Setting the mill volumetric load setpoint high (> 35% v/v) to reduce

impact breakage and promote abrasion, attrition grinding. Feeding the mill at feedrates > 1500 tph, however this had

detrimental affects on rougher flotation performance and causeddowntime due to downstream processing limitations (normally thetailings thickener).

Modification of the hydrocyclone underflow launder to allowhydrocyclone underflow to be recycled back to the SAG mill feed.

Installation of a limited variable speed capability for the SAG mill.(Installed December 1998).

Operation of the SAG mill with a low ball charge volume preventeddesign tonnages being achieved when the mill was fed with harder, morecompetent material.

The variable speed drive capability gives the operational flexibility to slowthe mill down if the charge load begins to empty when treating softer ore,while still maintaining the correct ball load for the harder ores.

Variable speed control is possible between 80% and 95% of maximum

speed. This allows the mill to be varied between 7.9 and 9.4 rpm, andthen in fixed speed at 9.9 rpm.

III-24

Typical Sag Mill Operating Trends

0

200

400

600

800

1000

1200

7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00

Time

FeedRate,

BearingWeight,

&Scats(ton),Mill

WaterFlow(m

3/hr)

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

MillPower(k

W)

S ag Fe ed Ra te S ag Mi ll Be ar in g R at e S ca ts S ag Mi ll Wa te r F lo w S ag Mi ll Po we r

Figure 4: Variation in SAG operation

The variable speed drive consists of an energy dissipation system,using the liquid resistance starters and an electrolyte cooling system.

Ore variability has decreased due to the mined ore becoming moreprimary in nature.

GRINDING MEDIA SIZE

SAG mill grinding media has been varied based on the top size ofmaterial reporting to the mill. During the later half of 1998, the gyratorycrusher gap increased due to accelerated wear of the crusher mantleand concaves. Coarse rocks up to 500mm in one dimension were

typical of feed reporting to the SAG mill during this time.

Tight maintenance and production schedules, and the lead time inmobilising equipment and personnel for the change out precluded animmediate mantle / concave replacement.

To soften the impact of decreased SAG mill throughput due to theincreased feed top size, larger grinding media were introduced, toimprove the impact breakage capability of the SAG mill. 105mm,125mm, and finally 140mm balls replaced the initial ball size of 90mm.The 140mm balls were added with some caution, as they represent a40% increase in ball weight over the 125mm balls, and consequently an

increase in potential liner damage due to ball impacts.

8/13/2019 05708 mill

8/16

III-25

105-mm balls have been charged to the mill since the crusher rebuildand SAG mill liner change (railbar replaced by tophat liners).

The ball size selection challenge for Ernest Henry is to successfullymarry ball breakage effectiveness with ball breakage frequency. Thisinvolves determining the smallest ball, capable of breaking thelargest/hardest rock, and then maximising their number in the ball

charge.

The ball mill was commissioned with 65mm balls. Grinding circuitsurveys indicated that the hydrocyclone underflow 80% passing size wasapproximately 700 800 microns, significantly less than anticipatedduring design. The new ball size of 40mm was selected based onestablished empirical relationships. Recent grinding surveys will reveal ifthe change in ball size has had a positive effect on reducing thecirculating load. There has not been any negative effects on the ballmilling circuit.

BALL CHARGING / BALL CHARGE VOLUME

Balls are lifted to the top floor of the grinding building in 20 tonne half-height containers. A series of bins and air-operated jaws allows aweighed amount of balls to be periodically added to the mill. Daily balladditions are based on the previous shifts throughput and the historicalconsumption rate. Additional balls are added if the ball charge in the millis below target.

The target ball charge volume for the SAG mill is 8 12 % v/v. The ballcharge volume has at times varied outside this range due to changes inore grindability. The installation of the variable speed drive on the SAGmill has enabled the mill to be ground out safely to provide an exact

measure of the ball charge volume. This procedure will be repeated atregular intervals, usually after major shutdowns, to maintain accurateconsumption rate data. Balls are added to the ball mill in a similarfashion, however the ball mill power draw is used to infer the requiredamount of balls. The ball charge in the ball mill is run at 35% v/v, whichrequired the installation of a dam ring to prevent ball losses to the ballmill discharge sump. The most recent ball consumption data are shownin Table 7.

Table 7: Media consumption rates

Actual Design

kg/tonne feed kg/kWh kg/tonne feedSAG Mill 0.281 0.032 0.3Ball Mill 0.188 0.041 0.46

III-26

Hydrocyclone Operation

The cyclone cluster of 8 inclined Krebs 840mm (33) hydrocyclonesreceives combined ball mill and sub screen deck size SAG milldischarge material via a 20/18 Warman centrifugal pump running with a1700 kW variable speed motor and gearbox.

The hydrocyclone circuit was designed based on a circulating loadranging from 200% of new feed to a maximum of 350%. Thesevolumetric loads required a maximum of 4 cyclones in use with 2 onstand-by. Operation of the circuit has shown that the recirculating loadsare significantly greater than design at times exceeding 500% of newfeed requiring up to 7 cyclones.

The higher than expected circulating loads result from:

High magnetite content of the ore concentrating in the hydrocycloneunderflow.

Inefficient cut of the large diameter hydrocyclones.

Inability to add additional feed water to the ball mill to achieve theoptimal rheology

8/13/2019 05708 mill

9/16

III-27

SAG MILL LINER/LIFTER DESIGN

The evolution of the Ernest Henry SAG mill liners is diagrammaticallyshown in Figure 5. Table 8 contains the liner/lifter specifications.

The SAG mill was supplied with ME International railbar design liners asspecified in Table 8. The original liners cracked longitudinally along the

liner portion of the lifter after approximately 3 months of operation.Several cracked liner sections fell out exposing the mill shell, causingminor damage and significant downtime. Subsequent production wasinterrupted every 5 - 7 days for SAG mill internal inspections to identifyliners, which required immediate removal. During February, March and

April 1998 a total of 67 liners were replaced resulting in the loss of theoriginal Hi-Lo lifter configuration. The loss of the Hi-Lo pattern did nothave any discernible affect on SAG mill grinding performance.

Figure 5: Liner profile evolution

Table 8: SAG mill liner specifications

Supplier Liner Type Lifter Face Angle Lifter Height Liner Thickness Liner Configuration

MEI Railbar 9 deg 205 mm 75 mm Hi - LoMEI Railbar 9 deg 205 mm 90 mm Hi HiMEI Tophat 15 deg 225 mm 100 mm Hi Hi

Vaughan Tophat 21 deg 225 mm 100 mm Hi Hi

The liner breakage has been attributed to excessive steel ball impacts onthe liner surface, which are caused by balls being thrown across the mill,past the toe of the charge. Overthrow is directly related to the lifter faceangle and speed of mill rotation.

III-28

The solution to the problem of liner breakage at Ernest Henry Mininghas been found by concentrating on improving the release angle of theballs by adjusting the lifter face angle. The implementation of thevariable speed option had not been incorporated when this work beganhence varying the mill speed was not an available option.

A shortage of liners required a quick redesign of the existing railbar

design. The redesign addressed the result of the real problem, ratherthan the cause, by beefing up the liner portion of the lifter as can beseen in Figure 5. The redesign did improve the life of the linershowever these liners did begin to fail later in their operational life.

Subsequent design changes also shown in Figure 5 indicate theconservative approach taken to achieving an improved face angledesign. The most recent inspection indicated wear rates ofapproximately 1mm of lift per 27,000 tonnes, with no broken liners todate.

Liner design is still ongoing with the latest set incorporating an

increased face angle of 21 degrees. Computer trajectory simulatorshave been used to optimise the face angle design. Both MillTraj (JKTech) and Millsoft (Peri Engineering) have been used in the simulationwork. Figure 6 shows a MillTraj simulation output for the 21-degreeface angle Top hat lifters.

21o

18o

15o

Figure 6: Trajectory simulation output using 15, 21 and 18 degreeface angles.

8/13/2019 05708 mill

10/16

III-29

POWER DRAW

Both mills have achieved and at times exceeded their design maximumpower draws, indicating that the mills were correctly sized and designed.

The SAG mill power draw is dependent on the charge specific gravityand the total charge volume. The target charge volume ranges from 25

35 % v/v depending on ore type. At 8 12 % v/v ball charge volume,and 25 35% v/v total charge volume the SAG mill typically draws 10 11 MW. The operating work index for the mill is extremely oredependent and can range from as low as 7 kWh/t when treating softsupergene material to as high as 14kWh/t on the harder primary ores.The average SAG mill operating work index for the project to date is 8.9kWh/t.

Power usage for the ball mill is typically greater than 95% of the availablepower. The power draw is maintained at this level by daily grindingmedia addition. The average operating work index for the ball mill is 4.6kWh/t.

GRINDING CIRCUIT CONTROL

Ernest Henry Mining adopted a philosophy of installing the latest in on-line instrumentation to facilitate grinding circuit stabilising control, and toprovide the necessary control elements for expert supervisory control.Table 9 summarises the installed grinding circuit instrumentation.

The plant uses a Yokogawa Centum CS distributed Control System forall stabilising control actions. The expert system programmed using theGenysm G2 operating environment, exchanges data with the DCS via acommunication bridge. The expert system decides on appropriatesetpoint changes based on a series of rules, then outputs the setpoint

changes to the DCS via the communication bridge.

Table 9: Grinding Circuit Instrumentation

Unit/Process Stream Instrum entation

Reclaim Apron Feeders Variable speed, cameras installed on discharge chutes

Stockpile reclaim conveyor Variable speed, weightometer, camera on SAG feed chute,SAG mill

Load cells, power draw meter, sound monitoring

All process water additions FlowmetersSump Level indicationHydrocyclone feed pumps Speed, amps,

Hydrocyclones Pressure, feed density and flowrate using magnetic compensators.Particle size and density measurement on the hydrocyclone overflow

Scats conveyors / diverter chute Weightometer, camera on discharge chutesBall addition hoppers Load cellsBall mill Power draw

III-30

95% of SAG mill control actions are based on the relationship betweenthe SAG mill load and the target mill load. The remaining 5% of controlactions involve emergency situations such as mill overloads, ball millcircuit overloads, or equipment operating outside their safe limits.

The target mill load is chosen based on the feed ore type and therequired throughput rate. Lower loads are selected when the ore is

hard, to maximise impact breakage while higher loads are selectedwhen the ore is soft, to maximise the generation of flotation feed sizematerial in the SAG mill. Feed tonnage and the feed water ratio areadjusted to maintain the target mill load.

The expert system does select between the power draw and the loadfor determining its control actions, however the mill power draw rarelyexceeds its Hi-limit and consequently has little influence on the controlactions.

Control actions are initiated every 2 mins, which typically result in smallchanges of 0 25 tph. These small changes have no destabilising

affect on rougher flotation because:

they represent a small fraction of the total feedrate, the large circulating load in the hydrocyclone circuit has a large

surge capacity, effectively dampening any small feedrate changes,and

the 127m3rougher flotation cells and large rougher circuit residence

time combine to provide significant surge capacity

Ball mill expert control uses water addition to the hydrocyclone feedsump to achieve the target flotation feed size. Hydrocyclone pressure isalso maintained at target, by adjusting the number of hydrocyclones in

operation.

The expert system has operated at all times, except for initial startupperiods. The operator acceptance of the system is excellent, as it freesthe control room operator to concentrate on other areas of the plant.

The plant and the expert system were commissioned simultaneously.

JK SAG CHARGE

JKSAGCharge is used to determine the mill charge volume and the toeand shoulder angles of the SAG mill charge on-line. The software is

based on Morrells power model. (Morrell, 1996, Valery & Morrell 2001)This information in turn can be used to help ensure that the mill isalways operated at the optimum volume filling and speed.

8/13/2019 05708 mill

11/16

III-31

Currently, JKSAGCharge is successfully being used at Ernest Henrydisplaying to the operators online:

Mill charge volume in cubic metres Total charge weight in tonnes Percentage of the mill volume occupied by coarse ore and ball

charge

Angular displacement of shoulder position of the charge Angular displacement of toe position of the charge

JKSAGCharge is interfaced directly with the DCS via the PI system, datacan be viewed on the operators monitor and directed to the DCS tocontrol the mill.

EFFECT OF RUN OF MINE FRAGMENTATION IN SAG MILLTHROUGHPUT

Ernest Henry mine is involved in a project with the JKMRC to Optimisethe Run of Mine Fragmentation for Downstream Processing. As a part of

this project the JKMRC team visited Ernest Henry mining and millingoperations and monitored one blast #2014-003 in Brecciated FelsicVolcanics (FV2).

The geological and rock properties, explosive properties and blast designparameters for this blast were collected. Fragmentation photographswere taken at the post blast muckpiles, at the back of trucks, at thecrusher product conveyor and at the SAG mill feed conveyor. Rocksamples were collected from the muckpiles to obtain rock mechanics andcomminution properties.

The data were used to calibrate the JKMRC fragmentation and

comminution models. These models were then used to simulate theeffect of blast design changes on milling performance.The following approach was adopted:

1. Rock characterisation from the existing geological data and from therock testing of samples.2. Fragmentation measurement of ROM, crusher product and SAG millfeed.3. Modelling ROM fragmentation and comparing it with the estimatesfrom Split.4. Simulating the impact of blast design changes on mill performance.

The simulations were performed for only one rock type ie. BrecciatedFelsic Volcanics (FV2).

III-32

BLAST MONITORING

Due to drill scheduling problems a part of blast #2014-003 was drilledwith 311mm diameter blast holes and the rest was drilled with 270mmblast holes. A major portion of the rock in that blast was identified asBrecciated Felsic Volcanics (FV2).

The structural data of the bench was collected by scan line mapping thebench face. Probability of fracture frequency estimated from scan linedata is shown in Figure 7. Rock samples were collected from each partof the blast (ie with 311mm holes and with 270mm holes) and weretested to determine the rock mechanic properties (Table 10).

0%

5%

10%

15%

20%

25%

30%

35%

0.5 1 2 3 4 5 6 7 8 9 10 11

Joints /m

Proba

bility

Figure 7: Fracture frequency observed on blast #2014-003 face

Table 10. Rock properties

270mm 311 mm Average

Density g/cc 2.97 2.93 2.95

Fracture toughness Mpa m0.5 0.61 0.67 0.64

Tensile strength Mpa 12 14 13

Young's modulus Gpa 56.2 75.4 65.8

Poisson's Ratio 0.3 0.27 0.29

UCS (direct) Mpa 189 207 198

Point Load Index 7.49 7.16 7.33

UCS (point load) Mpa 165 158 161

8/13/2019 05708 mill

12/16

III-33

The structural information from the scan line mapping (see Figure 7) andthe average rock properties from the Table 10 were used in the model toestimate the ROM fragmentation of blast #2014-003.

The ore from the two sections of the blast (270 and 311 mm drill holes)was fed to the primary crusher and the location of each truck was notedwhile it was dumping into the crusher. Photographs were taken at the

back of twenty trucks (10 from each section of the blast) to estimate theROM size distribution. At the same time photographs were also taken atthe crusher product conveyor. The time in which the ore remained in thestockpile was estimated based on the stockpile level and crusher productrate and SAG mill feed rate. Images from the SAG mill feed conveyorbelt were also taken. These images (photographs and video tapes) wereanalysed using the Split image analysis software to estimate the sizedistribution.

FRAGMENTATION MODELLING

Blasting researchers have used different approaches to estimate the

ROM fragmentation resulted from blasting. A detail review of history andstatus of fragmentation modelling is given by Kleine (1988), Sarma(1994) and Scott et al (1993). The Kuz-Ram empirical model is the mostpopular of all the models. However, the experience of JKMRC researchdemonstrates that the conventional Kuz-Ram model has the followingdrawbacks:

The rock factor used in the conventional Kuz-Ram model is quitesubjective though Lilly (1986) greatly improved on this.

The energy factor used in the Kuz-Ram model is derived from theideal detonation codes and does not take into account the effect ofconfinement and rock mass properties.

This model underestimates the percentage of fines (Kojovic etal.1995, Comeau 1996). In many mining applications this may not besignificant because the relevant size fractions in a blasted muck pileare relatively coarser. However, in softer ore types and in operationswhere the fines percentage affects the economics of the operation, itwill be important to estimate the fines with reasonable accuracy.

At the JKMRC, a fragmentation model has been developed based on anapproach similar to the Kuz-Ram but by improving on the drawbacksmentioned above.

Blast design parameters and explosive properties for the blast #2014-

003 are given in Table 11 shown on the next page.

III-34

Table 11. Blast design parameters and explosive properties forblast #2014-003

Holediameter mm 270 311

Burdenm 6.9 7.4

Spacingm 8 8.5

Holedepthm 17.5 17.5

Benchheight m 15.9 15.9Stemminglengthm 5 6.2

Chargelengthm 12 11.3

Chargeweight kg 930 1075

Powder factor Kg/m 1.06 1.07

Explosive Emulsion Emulsion

Density g/cc 1.25 1.25

VODm/s 4500 4500

RWS 95 95

The powder factor distribution of the blast is analysed using the JKsoftware 2D bench and showed that the energy distribution in the blastis not uniform and there is wide variation in the energy in the blast.

The ROM size distribution was estimated for the two sides in the blastby using the JK model and the results were compared with the sizedistributions estimated from the image analysis program Split (Figure 8).

0

10

20

30

40

50

60

70

80

90

100

1 10 100 1000 10000

Size mm

Cum%passin

g

Figure 8b: Comparison of ROM estimated by model and image

analysis (with 311 blast holes)

8/13/2019 05708 mill

13/16

III-35

0

10

20

30

40

50

60

70

80

90

100

1 10 100 1000 10000

Size mm

Cu

m%passing

JK Model

Kuz-Ram

Split

Figure 8b: Comparison of ROM estimated by model and imageanalysis (with 311 blast holes)

It is very well accepted that all image analysis programs including Splitunder estimates the fines and therefore needs site calibration. Such a

calibration can be done easily for images taken on conveyor belts but itmay be very difficult for ROM unless a sizeable quantity of material (atleast one or two truck loads) is sized. That is why the ROM sizedistribution was estimated from two sources of photographs, coarse endfrom truck images and fine end from crusher product images.

It is assumed that primary crusher contributes very little to the fines end(less than 20mm) of the ROM size distribution but reduces the coarseend of the ROM size distribution significantly. Therefore the fines end isestimated from the images taken from the crusher product conveyor andmerged with the coarse end estimated from the muckpile images andtruck images to get the full ROM size distribution.

The comparison of ROM estimated by the JK Fragmentation model andthe Split clearly shows that the ROM estimated by the model matcheswith the ROM size distributions estimated by the image analysis. It canalso be seen that the conventional Kuz-Ram model underestimates thefines significantly.

COMMINUTION MODELLING

Detailed survey data were collected during grinding circuit surveys of theErnest Henry mill. The gyratory crusher model was obtained using ROMand crusher product size distributions obtained by the Split system fromimages of muckpiles, trucks and conveyor belt collected at Ernest Henry.

SAG mill feed size distribution was also determined by sieving aconveyor belt cut of 10 metres using the technique developed by theJKMRC.

III-36

The SAG mill and screen model parameters were determined bynumerical non-linear least-squares fitting. The model parameters wereadjusted until the model best reproduced the observed product asclosely as possible from the corresponding feed to the unit and theoperating conditions. This process also provided estimates of data fitaccuracy and parameter estimation accuracy. All survey data sets fromthe SAG mills were fitted and the fit to the data was very good.

SIMULATIONS

Having validated the fragmentation model using the current blastdesign, the crushing and milling models were linked together to studythe influence of blast design changes on mill performance. The blast#2014-003 with 270mm diameter blastholes was chosen as the currentbench mark design and different alternate designs were simulated.Design 1 uses high shock energy explosive (with 5500 m/s VOD) andDesign 2 uses high shock energy explosive coupled with high powderfactor. The details of the blast designs are given in Table 12 and theestimated size distributions from the three designs are given in Figure 9.

Table 12. Blast designs used in the simulation

270mm(standard)

270-2 270-3 311mm(standard)

311-2 311-3

blast #2014-003 blast #2014-003

Bench height m 16 16 16 16 16 16

Burden m 6.9 6 5.5 7.4 6.5 6

Spacing m 8 7.5 6 8.5 7.7 7.2

Hole depth (L) 18 17.5 17 17.5 17.5 17

Explosive weight (Qe) 930 859 823 1,073 1,045 1,045

Stemming length 5 5.5 5.5 6.2 6.5 6

Subdrill 2 1.5 1 1.5 1.5 1

Explosive Emulsion Emulsion Emulsion Emulsion Emulsion Emulsion

Explosive densityg/cc

1.25 1.25 1.25 1.25 1.25 1.25

VOD m/s 4500 4500 4500 4500 4500 4500

Powder factor kg/m3 1.05 1.19 1.56 1.07 1.30 1.51

Drilling cost /m $11.42 $11.42 $11.42 13.63 13.63 13.63

Explosive cost /kg $0.90 $0.90 $0.90 $0.90 $0.90 $0.90

Stemming cost $23.10 $23.10 $23.10 $23.10 $23.10 $23.10

Service cost $126.72 $126.72 $126.72 $126.72 $126.72 $126.72

Drilling and blastingcost /t

$0.46 $0.53 $0.70 $0.46 $0.57 $0.65

8/13/2019 05708 mill

14/16

III-37

0

10

20

30

40

50

60

70

80

90

100

1 10 100 1000

Sizemm

Cum

%passing

270 - standard

270-2

270-3

0

10

20

30

40

50

60

70

80

90

100

1 10 100 1000

Sizemm

Cum%p

assing

311- standard

311-2

311-3

Figure 9: ROM estimates for different blast designs

Comparison between maximum mill throughput and SAG mill feed sizeobserved from blast #2014-003 (270-standard and 311 standard) andsimulated for different blast designs (270-2, 270-3, 311-2 and 311-3) are

shown in Table 13 (next page). The constraints for mill in designs 1 and2 are kept more conservative than the constraints in the blast #2014-003.

III-38

Table 13. SAG mill parameters and performance for three designs

Standard Blast#2014-003

Design270-2

Design270-3

Standard Blast#2014-003

Design311-2

Design311-3

Blast drill hole (mm) 270 270 270 311 311 311

Primary crusher gap 130 115 115 130 115 115

SAG feed rate (t/h) 1410 1530 1580 1410 1540 1570

Ball charge (%) 10 10 10 10 10 10

Total charge (%) 23 23 23 23 23 23

F80(mm) 120 109 107 119 109 107

F50(mm) 69 60 56 68 58 58

F20(mm) 12 10 8 12 9 8

%-50+20mm 15.8 17.0 16.9 15.8 17.1 16.8

%-10mm 18.6 20.0 21.4 18.7 20.7 21.4

It can be seen from Table 13 that the SAG mill feed from the differentblast designs resulted in similar quantities of material in the intermediateor so-called critical size material (between 20 and 50 mm). It can alsobe seen that the amount of fines (below 10.0 mm) in the feed increasedfrom 18.6% (blast 2014-003) to 21.4 % in blast designs 270-3 & 311-3.

It is important to note that it was possible to close the primary crushergap (close side setting) from 130 to 115mm and yet keeping similarthroughput and power draw for blast designs 270-3 and 311-3.However, in case of blast designs 270-2 and 311-2 the simulatedcrusher power was higher. The combination of blast design and closercrusher gap was responsible to achieve higher throughput rates throughthe SAG mill due to more favourable feed size distributions. It is well

known that the grinding rates and consequently throughput aredramatically affected by fines being carried through the mill (free grindmaterial that passes straight through the grates and trommel) and alsoby material in the top size and intermediate size. The simulations showthat Designs 270-3 and 311-3 increase the mill throughput by 11-12%compared to blast #2014-003.

As shown in previous studies conducted by the JKMRC, for a giventhroughput the minimum power draw and load volume is achieved witha relatively high F80, low F20 and a steep inclination of the particle sizedistribution (minimum amount of intermediate size material).

Such conditions are met with a bi-modal size distribution in which themajority of the feed resides in the coarsest and finest size fractions.

8/13/2019 05708 mill

15/16

III-39

This type of size distribution is likely to result in the minimum power drawand load volume and hence provide the lowest specific power andmaximum throughput capacity.

However, it is usually found that further increasing the F80 decreases themaximum throughput. This trend is emphasized with SAG mills runningwith a relatively high ball charge as is the case at Ernest Henry. It is

believed that this is due to the fact that the large ball charge provides thebulk of the grinding media rather than the rock charge. Coarser feedsizes therefore, have minimal impact on the grinding media but doesprovide a more difficult feed to grind. The desired feed size at ErnestHenry with primary ore seems to be an F80 of approx. 100-110 mm withmaximum possible amount of fines (% passing 10 mm) and minimum ofintermediate size fraction. With closing the primary crusher gap to 115mm this target F80 was achieved, but the rest of the curve depend onthe blast design as indicated during the simulations.

In this study the effect of blast designs 270-3 and 311-3 on muckpilediggability, back break and ground vibrations is not estimated. It is

essential to take into account these factors while performing any designchanges.

VALIDATION OF MINE-MILL BLAST DESIGNS

As the model simulations suggest significant improvements in millingefficiency by altering the blast design the next logical step is toimplement the best design suggested by the simulations in the field andvalidate it. A trial blast is proposed at Ernest Henry wherein one half ofthe pattern is designed with the standard designed and the other half isdesigned with 270-3 or 311-3 pattern. Ore from the two sides of the blastwill be fed to the mill in a controlled manner and the mill throughput will

be monitored. Some of the key factors for designing and monitoringsuch a trial blast are given below:

1. The trial blast block should contain sufficient ore to feed the mill for asufficient period of time (at least two days from each pattern) so that themill reaches stable operation.

2. Ore in the trial blast should have similar rock mass properties (ie onstandard side of the pattern and on modified side).

3. Mill should get at least 80% of the feed from one side of the blast (ieeither modified pattern or standard). It is advisable to avoid blendingwhile feeding the ore from trial blast.

The following parameters will be monitored during the trial blast:

III-40

Pre-blast:

Collect samples of rock from the trial blast for laboratory testing. Collect the drill penetration rates for each blast hole. Collect detailed information on the proposed blast design(s). Check the face and bench conditions.

During Blasts:

Check the implementation of the blast designs in the field. Conduct high speed photography or videography of the blast. Measure the velocity of detonation (VOD) in 1 or 2 holes in each

blast pattern.

Post blast:

Mining

Assess blast fragmentation through image analysis. Photographs

will be taken of the muckpiles and the back of trucks for both thesections of the trial blast. Survey post blast muckpiles to assess the throw and muckpile

swell. Measure the back break. Monitor the shovel performance.

Crushing

Monitor the crusher performance (throughput and power draw) andgap settings while feeding the trial blast ore.

Assess the fragmentation of crusher product. Photographs and/or

video of the crusher product conveyor will be taken while feedingthe ore from trial blast. These images will be analysed using Split toestimate the size distribution.

Grinding

Grinding circuit performance will be assessed by conducting surveysand mass balances of the circuit will be obtained.

Since a number of parameters will be monitored during these trials,communication between various groups is very important so thateveryone knows what to do and when.

Based on the simulations, Design 311-3 appears to be the optimumblast design in terms of maximising SAG mill throughput.

8/13/2019 05708 mill

16/16

III-41

CONCLUSIONS FROM THE MODELLING AND SIMULATION STUDY

Design of a full scale grinding circuit from bench scale testwork has beensuccessful at Ernest Henry Mining

One production blast #2014-003 in Brecciated Felsic Volcanics wasmonitored and the rock samples were collected from the blasted

muckpile to determine its rock mechanic and comminution properties.

The ROM size distribution for blast #2014-003 was estimated by usingthe image analysis program Split. The full size distribution was obtainedby merging the results obtained from truck images and crusher productconveyor images.

The ROM size distribution predicted by the JK Fragmentation model wasfound to be in close agreement with the size distribution estimated by theSplit.

The crushing and grinding simulations show potential to increase mill

throughput to 12% by altering the blast designs and primary crusher gap.

ACKNOWLEDGEMENTS

The Authors wish to thank Ernest Henry Mining management for theirpermission to publish this paper.

REFERENCES

1.A.R. MacPherson Consultants Ltd., Proposed Grinding System for the ErnestHenry Project for MIM Holdings Limited based on small scale tests at HazenResearch Inc. and Lakefield Research. April 1994.2.A.R. MacPherson Consultants Ltd., Proposed Grinding System for the Ernest

Henry Project for MIM Holdings Limited based on small scale tests at HazenResearch Inc. and Lakefield Research Addendum. August 1994.3. AMDEL, Ernest Henry Comminution Tests. AMDEL Report G768700G/94.May 1994.4.Brindley, S. et al. (1999) Case Study Ernest Henry Mining: Experiences andlessons to be learned from initial grinding circuit design to the end of the first yearof operation. Crushing & Grinding 99, IRR Conference, Perth 29-31 March.5.Brown P, 1994 - 1996, Met Engineers, Mineralogical analysis and flotationtestwork on Ernest Henry ore samples, Reports KM433, KM460.6. Comeau, W (1996), Explosive energy partitioning and fragment sizemeasurement Importance of correct evaluation of fines in blasted rock, Proc. ofthe Fragblast-5 workshop on Measurement of Blast Fragmentation, Montreal237-240.

7. Cunningham, C.V.B. (1983), The Kuz-Ram model for prediction offragmentation from blasting, Proc. 1st Int. Symp. on Rock Fragmentation byBlasting, Lulea, 439-454.

III-42

REFERENCES (Continued)

8.Ernest Henry Feasibility Study, January 1995.9. JKTech, Ernest Henry SAG Mill Trajectory Simulations. JKTech Job No.99013. January 1999.10. JKTech, Grinding Circuit Simulation Design Study for the Ernest HenryProject. May 1994.11.Kleine, T.H. (1988), A mathematical model of rock breakage by blasting,

Ph.D Thesis, Univ. of Queensland, Australia.12. Kojovic, T, Michaux, S and McKenzie, C (1995), Impact of blastfragmentation on crushing and screeing operations in quarrying, Explo 95,Brisbane, pp 427-436.13.Lilly, P.A. (1986), An empirical method of assessing rock mass blastability,The Aus. I.M.M./I.E. Aust. Newman Combined Group, Proc. Large Open PitMining Conference, pp. 89-92.14. Morrell, S. 1996 The Prediction of Grinding Mill Power. TransIMM, 101,C25-3215. Nelson, M., Valery Jnr., W., and Morrell, S. (1996) Performancecharacteristics and optimisation of the Fimiston (KCGM) SAG mill circuit.International Conference on Autogenous and Semiautogenous GrindingTechnology, 1, p. 233-248, Vancouver, Canada.16.Orway Mineral Consultants, Assessment of Grinding Characteristics fromDrill Core Samples. June 1994.17.Ryan A.J, 1998. Ernest Henry copper-gold deposit, in Geology of Australianand Papua New Guinean Mineral Deposits (Eds: D A Berkman and D HMackenzie), pp 759-768 (The Australasian Institute of Mining and Metallurgy:Melbourne).18. Sarma, K S (1994), Models for assessing the blasting performance ofexplosives, Ph.D Thesis, Univ. of Queensland, Australia.19. Scott, A., Chitombo, G and Kleine, T. (1993), The challenge of theprediction and control of fragmentation in mining, Proc. 4th Int. Symp. on RockFragmentation by Blasting, ed. Rossmanith, Balkema, Rotterdam, 507-517.20. Strohmayr S.J., Barns K.E., Brindley S.K., and Munro P.D., 1998 Mineralogy Controlling Metallurgy at Ernest Henry Mining., in Proceedings ofthe Mine to Mill 1998 Conference, Brisbane pp 13-18 (The Australasian Institute

of Mining and Metallurgy).21. Tew, A.F. et al 1999, Worlds Largest Shell Supported SAG Mill at theErnest Henry Copper Concentrator Operating Experiences and Results.SMME Annual Meeting, Denver, March 1999.22. Valery, W and Morrell,S. 2001 JKSAGCharge, A System to EstimateCharge Volume and Position Online in AG and SAG mills. Comminution, 2001

Conference 21-23rd

March 2001. Brisbane 2001.