Improved characterisation of ball milling energy ...623483/UQ623483_OA.pdf · This paper describes...

Post-print copy

Please reference as: Ballantyne, G.R., Hilden, M. and van der Meer, F., 2017. Improved characterisation of ball milling energy requirements for HPGR products. Minerals Engineering (accepted 8 Jun 2017).

Improved characterisation of ball milling energy requirements for HPGR products Grant R Ballantyne1*, Marko Hilden1 and Frank van der Meer2 12 1The University of Queensland, Sustainable Minerals Institute, Julius Kruttschnitt Mineral Research

Centre

2Weir Minerals

(*Corresponding author: [email protected])

Accepted for publication in: Minerals Engineering

Published version found online at: http://www.sciencedirect.com

Abstract This paper describes a method for assessing the downstream milling energy requirements for high

pressure grinding rolls (HPGR) products based on a Bond mill test procedure. Multiple trade-off studies

have reported the performance of HPGR versus SAG milling with energy savings of between 11 and

32 per cent. One factor that is often inconsistently defined in these studies is the change in the Bond

Ball Work Index (BBWi). The Bond test can overestimate the reduction in ball milling energy

requirements for HPGR products, not due to a change in the breakage characteristics of the particles,

but because the Bond test feed for a sample crushed in a HPGR has a greater fines content than a

conventionally-crushed sample. This paper rigorously assesses the actual change in BBWi achieved

through the use of HPGR technology.

The reduction in BBWi was found to not be dependent on the ore hardness and thus the expression

of the change as a per cent is incorrect. When the size distribution of the HPGR product was matched

to the crushed product, more than 95% of the samples tested resulted in a reduction in Work Index,

with an average reduction of 1.9 kWh/t. A proportional reduction was seen when the Size Specific

Energy (SSE) was calculated for the reconstituted samples. Six samples were tested where the original

size distributions were retained and these saw greater reduction in BBWi than SSE.

A gold mine was surveyed to investigate the ball milling requirements of an industrial HPGR circuit.

Two surveys of the ball mill operating at different conditions found that milling efficiency could be

improved by 20%. These results highlight the importance of the mills operating conditions on energy

efficiency as they can overwhelm the potential benefits of HPGR pre-conditioning.

1 Comparative Bond work index data were supplied by Frank van der Meer, Evert Lessing and Ric Stocco of Weir Minerals. 2 Design, laboratory and operating data from Tropicana Mine were supplied by Nick Clarke, Mike Di Trento of AngloGold Ashanti, and Fred Kock and Brian Putland of Orway Mineral Consultants (OMC)

mailto:[email protected]

http://www.sciencedirect.com/

Post-print copy


Keywords HPGR; Bond; Work Index; Comminution; Energy

Highlights 1. The ball milling requirements of HPGR and standard crusher products are measured

2. Methods for assessing the energy consumption of the Bond Ball Mill are discussed

3. The average reduction in Work Index of HPGR products was 1.9 kWh/t

4. A similar magnitude of reduction was observed in the Size Specific Energy

5. Ball milling efficiency can vary significantly independently of HPGR operation

Graphical Abstract Relationship between reduction in BBWi and measured SSE75 with HPGR feed preparation instead of

staged crushing. Circled data points are from tests where the natural size distributions were used and

the two size distributions from crushing and HPGR were not matched.

Introduction The introduction of high pressure grinding rolls (HPGRs) to the minerals industry has not led to it

overtaking the well-established semi-autogenous grinding (SAG) technology despite the widely

Reconstituted

Natural PSD

-3

-1

1

3

5

7

9

11

-1 0 1 2 3 4 5 6

Red

uct

ion

in m

easu

red

SSE

75

(kW

h/t

-75

μm

)

Reduction in BBWi (kWh/t)

Difference in ball millling energy for HPGR products

Post-print copy


reported efficiency benefits. Multiple trade-off studies comparing different comminution circuit

options have estimated that HPGR circuits can offer energy savings of between 11 and 32 per cent

compared with SAG milling circuits (Davaanyam et al., 2015). The confined-bed breakage mechanism

employed by the HPGR requires less energy to achieve the same degree of size reduction as SAG

milling. The cost of the grinding media consumed in SAG mills provides an additional benefit that is

greater for more competent ores. However, the focus of this paper is the reduced energy

requirements of ball milling circuits following HPGRs. This is typically quantified based on Bond ball

work index (BBWi) tests of HPGR products, which generally report a significantly lower BBWi value

than for the HPGR feed (Wang, Nadolski et al 2013). The implications for the equipment sizing of the

downstream ball mills can be significant (Patzelt et al., 2006). Schönert (1988) estimated that replacing

existing circuits with HPGR-ball mill circuits could increase circuit capacity by 12-25%, and reduce

energy by 10-20%. However, for this to be achieved the HPGR must produce at least 10% of the final

circuit product and, furthermore, damage those particles remaining unbroken.

The reduction in BBWi has been shown (Baum et al., 1997; Daniel, 2007; Esna-Ashari and Kellerwessel,

1988; Otte, 1988), at least in part, to be due to the appearance of “microcracking” fracturing of

particles in the HPGR. Daniel (2007) observed microcracks in HPGR product using mineral liberation

analysis while Lin et al. (2012) measured the specific internal surface area contained within them using

tomography, but their contribution to reductions in BBWi has not been directly quantified.

Tavares (2005) found that HPGR product particles coarser than 1.5 mm required on average 35% less

energy to fracture in comparison to crushing and produced finer progeny when subjected to single

particle impact crushing. This weakening was found to increase with compaction pressure and was

independent of particle position within the confined bed, but was not observed for particles finer than

1.5 mm. Tavares (2005) used this methodology to decouple the effects of increased fines production

and particle weakening on the ball mill through simulation of multiple sequential breakage steps. Shi

et al. (2006) also found that the increased grindability of HPGR products was greater when coarser

closing screen sizes were used in the Bond Ball Mill test. van der Meer and Schnabel (1997) used pilot

plant tests, torque mills of different sizes, a Hardgrove mill as well as standard Bond Ball mill tests to

show that the grindability of HPGR products were reduced in comparison to crushed products, and

that this reduction increased with applied pressure. Watson and Brooks (1994) supported the finding

that the Work Index reduced linearly with increasing pressing force. Stephenson (1997) also found

that the reduction in competence was ore-dependent, and either related to the rock structure

(presence of vesicles reduced the degree of microcracks) or the fracture toughness (reduction in work

index was only observed for ores with high fracture toughness). Morrell (2009) suggested that a

reduction in work index of 5% could be assumed in the absence of full scale data.

There is some evidence that microcracks could lead to increased liberation if the mineral association

promotes cracks following the inter-species grain boundaries (Battersby et al., 1992). The evidence

for this is inconclusive and Daniel (2007) noted that accurate quantification of the effect of breakage

mechanism on liberation as measured using SEM-based liberation analysis is not straightforward.

Clarke and Wills (1989) compared tin ore crushed using a rod mill and HPGR and found that liberation

was enhanced by compression breakage. Studies by Patzelt and Knecht (1996) and Baum and Ausburn

(2011) for example have compared leachability of ores from HPGR and crushers finding that high-

pressure rolls have a higher leaching rate, in particular in the coarsest size fractions. Additionally, Shi

et al. (2006) found that HPGR treated material showed increased flotation responses. In contrast

Post-print copy


however, neither Palm et al (2010) nor Solomon et al (2011) found evidence that HPGR offered any

preferential breakage or increased flotation recovery for South African PGM ores. Kodali et al (2011)

found no increase in leaching kinetics for a sulphide copper ore but some improvement for an oxide

copper ore. Vizcarra et al. (2010) did not find any increase in liberation between compression and

impact breakage and Garcia et al. (2009) and Xu et al. (2013) observed increased liberation only at

impractical compression rates of 0.5 cm/day or slower. Therefore, although HPGR use compression

breakage and increased liberation can be achieved via this breakage mechanism, the rate required for

high throughput may be too fast for this to be realised in practice.

The Bond Ball Mill Work index (BBWi) is the industry standard procedure for assessing the grindability

of ores. The test involves conducting a locked-cycle grinding test using Bond’s standard mill design.

The locked-cycle test enables a batch process to emulate the grinding behaviour of a continuous mill

with a 250% recirculating load. The mill product from each cycle is screened and the oversize is

combined with fresh feed (of the same mass as the screen undersize which is removed) to maintain a

constant filling. This process is repeated for as many cycles as is required to obtain a constant

production rate of screen undersize. The standard feed for the test is prepared through staged

crushing a sample to 100% passing a 3.35 mm sieve. Bond (1961) defines his work index according to:

BBWi = 4.91/[P10.23 Gpr0.82 (P80

-0.5 – F80-0.5)] – Equation 1;

where P1 is the closing screen opening, in microns; P80 and F80 are the test product and feed 80%

passing sizes, respectively, in microns; Gpr is the net grams per revolution averaged from the last three

cycles. The numerator in Equation 1 of 4.45 published in the original equation corresponds to the

work index for a short ton, whereas 4.91 should be used when using the metric tonnes.

An implicit requirement of the Bond test is that its feed and product size distributions should be, at

least approximately, parallel (Musa and Morrison, 2009). The product from a HPGR (and SAG mills)

tend to contain a higher proportion of fine material than a crushed feed specified for the standard

Bond test (van der Meer and Gruendken, 2010). However, even with significant differences in the

feed size distributions, the product size distributions from the Bond tests have been found to be

identical and solely controlled by the closing screen aperture (Shi et al., 2006). Hence the Bond ball

mill test should be modified when applied to the grindability of HPGR product. Figure 1 shows the

product of progressive crushing to 100% -3.35 mm (standard Bond procedure) and the product of a

HPGR locked-cycle test with a closing screen of 3.35 mm. Although the 80% passing sizes are similar,

in this example, there was 14% material finer than 75 µm in the HPGR product compared with 8% in

the crusher product, but the difference can be greater in some circumstances. The Bond test

procedure can account for differences in the quantity of fines in the feed by calculating the net grams

per revolution produced. And although Rowland and Kjos (1980) developed many correction factors,

none of these relate to the degree of fines in the feed. Amelunxen and Meadow (2011) discussed a

methodology for modifying the BBWi by introducing a net correction factor (CFnet) to allow for

different degree of fines with HPGR, SAG or crushed products, but the method remains proprietary.

There also exists a number of indirect methods that are used by practitioners such as the ‘phantom’

cyclone approach developed at the Julius Kruttschnitt Mineral Research Centre (JKMRC) and

functional performance analysis (Bartholomew, et. al., 2014). Because of its wide acceptability in the

industry, the Bond ball mill test is the most pragmatic choice for investigating the ball milling

requirements of HPGR products, but it may not be the most accurate. This study aims to separate the

Post-print copy


pre-weakening (or microcracking) effect of HPGR from the fines production in the Bond ball mill test

results.

Figure 1 - Size distributions of the product from HPGR locked cycle test with 3.35 mm screen and staged crushing to 100% -3.35 mm using the same ore.

Methods Two approaches are described in the literature to reduce the influence of extra fines for determining

the BBWi of HPGR products and isolate any potential pre-weakening effect. (Stephenson, 1997)

described reconstituting the HPGR product to match the size distribution of the staged crush size

distribution. This approach will be explored further in this paper. Alternatively, a full population

balance model can be developed of the Bond mill as per the (Rajamani and Herbst, 1984) method.

Two other approaches have avoided using the Bond work index altogether. Patzelt et al. (2005)

utilised a proprietary laboratory (dry, open-circuit) mill test that relates the generation of fines to the

energy input. And Tavares (2005) conducted single particle impact crushing to investigate the relative

strengths of the HPGR and crushed products.

Reconstitution approach

HPGR products typically have a high fines content that could cushion the grinding action in the Bond

test, reduce the grinding efficiency, and consequently increase the work index. The change in fineness

is also not captured by the F80 value that is used in Bond’s analysis, as this is more closely related to

the closing screen aperture chosen for the test. To eliminate the influence of feed size distribution,

comparison of the Work indices of the feed sample and the HPGR products has been screened into

between 6 and 10 narrow size fractions, for instance the fractions 2.8-3.35 mm, 2.0-2.8 mm, 1.0-

2.0 mm, 0.5-1.0 mm, 0.125-0.5 mm, and -0.125 mm. From this feed stock of size fractions, the initial

Bond test feed is reconstituted into the same size distribution as the conventionally crushed sample.

Staged crush

HPGR

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

10 100 1000

Cu

mm

ula

tive

pe

r ce

nt

pas

sin

g (%

)

Sieve size (µm)

Post-print copy


The same is performed for the required new feed additions between each Bond test cycle. Thus, the

sample for the determination of the Bond Work index of the HPGR products will have the same size

distribution as the samples of the conventionally crushed, HPGR feed.

In breakage processes, the more competent components in the feed tend to report to the coarser

fractions in the product size distribution (Bueno et al., 2013). Thus when preparing a Bond test feed

(which requires the material to be sized to minus 3.35 mm), scalping the crushed product without

recycling the oversize, might skew the mineralogical composition of the Bond test feed. Therefore,

the sample for Bond WI determination on the HPGR product was taken from closed circuit HPGR

product or laboratory locked-cycle tests, where minerals and particle sizes have equal opportunity to

be subjected to the HPGR process. In addition, in a closed circuit arrangement, an enrichment of the

harder components from the ore may take place in the recycle stream and consequently in the HPGR

feed. To ensure a properly prepared feed, only the steady state product of the closed circuit HPGR

crushing was used. This fraction provided a fully representative sample for the Bond procedure and

the results thereof can be considered as indicative of the final HPGR product. In using this technique

consideration should be made to the fact that in reconstituting the crusher product size distribution,

the coarse particles are enriched. The result of this will depend on whether the coarse particles are

more competent due to surviving the HPGR or weaker due to an increased degree of microcracking.

Size Specific Energy (SSE) approach

The Size Specific Energy (SSE) is defined as the specific energy required to generate new material finer

than a particular marker screen aperture (Ballantyne et al., 2015). The most common marker size is

75 µm, producing the SSE75 with the units kWh/t-75µm. The SSE75 is a useful measure for comparing

the grindability of the products from crushing and HPGR because it accounts for the increase in fines

in the feed and most of the surface area is contained in the -75um fraction. The SSE75 can either be

calculated from single particle impacts (Ballantyne et al., 2015) or from Bond test results (Levin, 1992).

The SSE75 results in this paper are obtained from Bond test data.

The calculation of SSE75 from the Bond ball test is similar to the approach taken by Patzelt et al. (2005).

That technique compared the grindability of two HPGR fractions (centre and total) and conventional

crushed products in a progressive grinding test in a dry, open-circuit laboratory ball mill (ibid). The

amount of fines (-90 µm) was measured in the feed and then subsequent to each progressive grinding

step along with the specific energy input to the mill. Figure 2 shows the results from two ore samples

that were presented by Patzelt et al. (2005) with the amount of fines in the feed subtracted from the

product to calculate the generation of new fines. The HPGR products had between 10 and 20% more

fines in the feed, and thus the conventional BBWi values were reduced by 10 to 20% (ibid). However,

when the generation of new fines was calculated, both the crushing and HPGR samples exhibited the

same competence in the ball mill. Ore 1 was less competent with a Size Specific Energy at 90 µm

(SSE90) of 18.3 kWh/t-90µm, whereas Ore 2 had a SSE90 of 22.9 kWh/t-90µm.

Post-print copy


Figure 2 - Generation of -90 µm material with increasing specific energy applied in a laboratory ball mill on the product from crushing and HPGR calculated from data published by Patzelt et al. (2005).

It is possible to calculate the SSE using the industrial standard test procedure of the Bond ball mill test,

however it requires a measurement of the energy consumption of the mill. The Bond test relies on

empirical fitting parameters determined by Bond and expressed as exponents on the closing screen

size (0.23) and the net grams per revolution (0.82). These parameters were fitted to equate the results

from the standard Bond Ball test to a standard pilot plant utilising a 2.44 m diameter ball mill (Bond,

1961). As highlighted by Levin (1989), the results from the Bond test can be used to calculate the

predicted power draw of the mill. The standard Bond (1952) equation can be used to calculate the

specific energy required to go from a specific feed particle size (F80) to a product size distribution (P80):

Specific energy (kWh/t) = 10 BBWI (P80-0.5 – F80

-0.5) – Equation 2.

Substituting equation 1 into equation 2 will result in a calculation for the specific energy requirements

of the Bond’s 2.44m pilot mill directly from the passing screen size (P1) and the grams per revolution

(Gpr):

Specific energy (kWh/t) = 49.1/(P10.23 x Gpr0.82) – Equation 3.

Equation 3 can be rearranged to provide the effective joules per revolution (J/rev) that can be

expected of the pilot mill:

Energy (J/rev) = (177Gpr0.18)/(P10.23) – Equation 4.

Levin (1989) used a similar (but slightly different) calculation and found that the average energy value

of a Bond test was 1425 x 10-6 kWh/min or 71 J/rev. However, analysing the same data using the

equation above resulted in an average figure of 62 J/rev. In using an average figure, Levin (1989)

assumed that although the energy consumption is dependent on both the grams per revolution and

the closing screen, these factors are related and thus there were ‘core (normal) values’ that could be

averaged. Values of energy consumption outside the defined ‘core (normal) values’ were thus

determined to be errant. However, when applied to a larger dataset, the values are found to be more

normally distributed with fewer errant values (see Figure 3).

0

10

20

30

40

50

60

70

80

90

0 5 10 15 20

Ge

ne

rati

on

of

ne

w -

90

µm

mat

eri

al (

%)

Specific Energy (kWh/t)

Crush Crush

HPGR HPGR

HPGR centre HPGR centre

Ore 1 Ore 2

Post-print copy


Figure 3 - Calculated values of energy using expanded JKMRC database (expanded from (Levin, 1989) – red bars).

In addition to expanding the ‘core (normal) values’ identified by Levin (1989) to a distribution, adding

the expanded database identified that the energy consumption was highly dependent on both closing

screen aperture and net grams per revolution (Figure 4). Although the net grams per revolution is

dependent on the closing screen size, this relationship is not sufficient to overcome the relationship

developed by Bond (1961).

Figure 4 - Relationship between closing screen aperture, net grams per revolution and energy consumption (solid lines represent the model equation and data points the values calculated from the actual Bond test).

Daniel (2007) measured the net electrical power draw of a laboratory Bond mill to be 91.4 J/rev. This

experimental value is comparable to 93 J/rev estimated by Bond (1949); 91.4 J/rev estimated using

the Morrell (1992) power model; and 93.2 J/rev estimated by the Hogg and Fuerstenau (1972) power

0

50

100

150

200

250

300

40 50 60 70 80 90

Fre

qu

en

cy in

dat

abas

e

Calculated values of E (J/rev) (For a pilot mill with 2.44m diameter)

GPR = 1

GPR = 2GPR = 3

0

10

20

30

40

50

60

70

80

90

10 100 1000

Ene

rgy

con

sum

pti

on

(J/

rev)

Closing screen aperture (µm)

Expanded database

Levin (1989)

Post-print copy


model. The power draw has also been measured using a torque meter for Bond tests conducted at

Weir’s KHD laboratories in Germany, yielding a mean of 81.5 J/rev and a standard deviation of 3.7

J/rev (see Figure 5). This energy requirement of the mill is not affected by external parameters such

as closing screen size or grindability unless the frictional characteristics, density or viscosity changes.

The effective power draw calculated using the Bond (1961) formula therefore encompasses the

inefficiencies of the laboratory mill in comparison to the pilot mill. Since the closing screen and net

grams per revolution is used to calculate the BBWi, there is a relationship between BBWi and the

effective energy efficiency of the laboratory mill (Figure 6). The lower energy efficiency of the

laboratory mill may also be due to running dry—Tüzün (2001) found that compared to regular dry

Bond tests, a wet Bond test used 30% fewer revolutions. The lower energy efficiency at higher BBWi

may also be a reflection of the lab-scale mill having insufficient energy to efficiently grind coarse

competent rocks.

Figure 5 – Calculated, estimated (Daniel, 2007) and measured energy consumption from Bond tests with the grams per revolution (g/rev) indicated by the width of the data points.

50

55

60

65

70

75

80

70 75 80 85 90 95 100

Ener

gy c

alcu

late

d f

or

a p

ilot

mill

wit

h 2

.44

m d

iam

eter

(J/

rev)

Energy measured via torque on laboratory mill (J/rev)

Width ~ g/rev

(Dan

iel,

20

07

)

Post-print copy


Figure 6 - Relationship between BBWi and the calculated energy efficiency of the laboratory mill from Bond's formulas.

A calculation or measurement of the energy consumption of the laboratory is necessary to calculate

the SSE75 from the Bond (1961) test. This investigation highlighted that using the Bond (1961)

calculated energy is flawed because it takes into account the inherent inefficiencies of the laboratory

test. However due to its widespread acceptability, it will be used as an initial estimate. The SSE75 can

be calculated using Equation 3 as a basis and including the generation of new –75μm:

SSE75 (kWh/t–75μm) = 49.1/[P10.23 Gpr0.82 (%–75μmprod - %–75μmfeed)] – Equation 5.

The actual power draw of the mill was also measured for selected tests using a torque sensor on an

instrumented ball mill. This measurement was used to obtain a ‘measured SSE’ which was compared

to the above ‘calculated SSE’.

The SSE75 has been calculated and measured from the database of Bond Ball mill tests using the

methodologies outlined previously. The calculated SSE75 compared well with the BBWi calculated

from the same test (Figure 7), although the SSE75 is a slightly greater value (as shown by the parity

line). The BBWi corresponds to the specific energy required to reduce intact rock structure to a P80 of

100 µm; whereas the SSE75 represents the energy required to produce a tonne of new minus 75 µm

material. Therefore, the difference between the two measures mathematically equates to the specific

energy required to reduce the material from a P80 of 100 µm to 100% passing 75 µm. In reality this is

not exactly true as the gradient of the particle size distributions plays an important role in these

calculations. The measured SSE75 was higher again than the calculated SSE75 as the measured energy

consumption was significantly greater than the calculated energy. Also, because the energy

consumption is measured, the energy does not vary with closing screen size or grindability, but with

the filling and speed of the mill.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 10 20 30 40 50

Bo

nd

en

erg

y e

ffic

ien

cy o

f la

b m

ill (

%)

Bond Work Index (kWh/t)

Torque measurement

Assumed 93 J/rev

Post-print copy


Figure 7 - Comparison of calculated and measured SSE75 and BBWi as assessed by the standard Bond Ball Mill Work Index laboratory test.

Results 200 Bond Ball Work Index (BBWi) tests were conducted at Weir’s KHD laboratories on products from

both HPGR and progressive crushing. The samples came from a range of commodities and countries:

32 gold, 73 copper, 67 iron and 28 from other commodity groups, and these occurred in 24 countries

globally. The HPGR products were reconstituted to be identical to the progressive crushing product

before conducting the BBWi test, and therefore any differences are likely to be due to a weakening

effect. As shown in Figure 8, the crushed product was consistently measured to have a higher BBWi

than the HPGR product. The repeatability of Bond tests has been measured by Weier (2016) to be

±4.8% (95% Confidence Interval) which is significantly smaller than the average difference. The

reduction in BBWi with HPGR processing was not related to the Work Index of the ore and thus the

standard method of representing the reduction as a percentage is not applicable (see Figure 8). The

average reduction in BBWi across all these tests was 1.7 kWh/t and a standard deviation of 1.1 with

no significant influence of the ore-types evaluated (copper, gold and iron ores). Stephenson (1997)

found the reduction to be dependent on ore-type, therefore a more detailed description of the gangue

mineralogy and grain structure may show some dependence. The distribution of BBWi from the tests

was converted into a box-and-whisker plot and the median was used to plot the reduction on the Bond

Energy Intensity Curve (Ballantyne and Powell, 2014). The distribution shown for both HPGR and crush

product do not represent the experimental error, but the distribution of the database results. The

position of the crushed ore near the 50th percentile is a reflection of the fact that the majority of the

underlying energy curve have been measured on crusher product. The reduction in BBWi from

crushed to HPGR product is shown to be significant as it results in an effective reduction from the 48th

percentile to the 20th percentile on this Energy Curve (Figure 9).

SSE75 Measured

SSE75 Calculated

0

5

10

15

20

25

30

35

40

45

50

0 5 10 15 20 25 30 35 40

Size

Sp

ecif

ic E

ner

gy (

kWh

/t-7

5μ

m)

Bond Work Index (kWh/t)

Post-print copy


Figure 8 - Reduction in BBWi with HPGR product using the reconstitution method.

Figure 9 - Bond work index Energy curve with box-and-whisker plots showing the distribution of BBWi for standard crushed product and HPGR product from the Weir database.

0

5

10

15

20

25

0 5 10 15 20 25

HP

GR

pro

du

ct B

BW

i (kW

h/t

)

Crushed product BBWi (kWh/t)

Post-print copy


The SSE75 calculation for the Bond tests in the Weir database were not expected to dramatically alter

the results seen from the BBWi because the HPGR product size distributions are matched to the

product from crushing. This result was confirmed when the reduction in BBWi was plotted against the

reduction in the measured SSE75 (see Figure 10). In addition to all the tests with matched size

distributions between HPGR and crushed products, six tests were conducted where Bond tests were

conducted on the natural products. These tests are highlighted with red circles in Figure 10. It was

observed that when the Bond test was conducted on the naturally finer HPGR products, the reduction

in BBWi was greater than the reduction of SSE75. The reduction in SSE75 was significantly lower in all

but one case and most tests actually showed an increase in competence when the HPGR was used as

the feed preparation. This increase is most probably a result of the additional fines cushioning the

impact events within the mill.

Figure 10 - Relationship between reduction in BBWi and measured SSE75 with HPGR feed preparation instead of staged crushing. Circled data points are from tests where the natural size distributions were used and the two size distributions from crushing and HPGR were not matched.

Discussion The Bond ball mill test is the industry standard in testing the grindability of ore, but it is dependent on

the ore being prepared via staged crushing. It can be used to assess the grindability of ore from

alternative processes such as HPGR, but this method comes with risks. The increased fines found in

the HPGR product has the potential to cushion impacts within the mill as well as providing more

material close to the product screen size in the feed. The reconstitution method utilised here enables

the size distributions to be matched, thus avoiding these problems. However, in so doing, it may

introduce alternative bias in the results. The HPGR shows the natural tendency to preferentially break

softer material. Thus, the coarser, more competent, material that is concentrated when the sample

is reconstituted to the crushing size distribution. This would have the effect of reducing the perceived

benefit of HPGR over crushing which is a lower risk than the alternative.

-3

-1

1

3

5

7

9

11

13

-1 0 1 2 3 4 5 6Red

uct

ion

in S

SE7

5 (

kWh

/t-7

5μ

m)

Reduction in BBWi (kWh/t)

Post-print copy


Industry implications

The reduction in BBWi following HPGR crushing has been presented in many industrial design case

studies (Amelunxen and Meadows, 2011; Oestreicher and Spollen, 2006; Wang et al., 2013). One such

example was Vanderbeek et al. (2006) who attributed a 10% reduction in Bond work index at Cerro

Verde to HPGR induced microcracking. However, no expectation of a weakening effect was assumed

in the design process for equipment sizing; the ball mills were sized based solely on the increased fines

expected from a SAG mill in comparison to a Cone crusher product. In a subsequent study at the Cerro

Verde site, Koski et al. (2011) found that microcracks were present in coarse HPGR product at high

pressures, however the ball mills required 20% more capacity than designed. This shortfall was

attributed to neither achieving the expected fines generation in the HPGR nor the measured reduction

in BBWi. Had the measured reduction in BBWi been accounted for in the design, the difference

between design and performance of the ball mills would have been 34%. Due to the commercial

importance of achieving design throughput capacity and the cost implications of under- (or over-)

designing a circuit, accurate determination of the ball milling requirements of HPGR product is

imperative.

Tropicana Gold Mine provides another case study in the ball milling requirements following HPGR.

Anglo Gold Ashanti’s Tropicana Gold Mine is located in the newly developed Albany-Fraser Orogen,

east of Kalgoorlie, Australia. Following an extensive feasibility study, a HPGR-ball milling circuit was

justified based on the reduced power requirements and reduced operating expenses relative to SAG

milling. The HPGR-ball mill circuit was estimated to require 75% of the energy of a standard Semi-

Autogenous Grinding (SAG) mill circuit. Even with an anticipated greater capital cost, the HPGR-ball

mill circuit was found to give significantly better financial returns in all cases investigated in the

feasibility study (Ballantyne et al., 2016). A comparison between the specific energy requirements of

the HPGR, crush and SAG based circuits (SABC) from the feasibility study is shown in Figure 11 on the

Tonne Intensity Energy Curve (Ballantyne and Powell, 2014). The specific energy requirements of all

options lie in the upper quartile of major operating mines in the sector, however the HPGR circuit is

considerably more efficient. This benefit combined with the high cost of electricity at the mine

motivated the decision to employ a HPGR circuit design. Analysis of the grindability of HPGR product

was conducted in the feasibility study and it was found that although a reduction was seen when the

natural size distributions were tested, when the reconstitution method was used, this reduction

disappeared. A series of timed grinds were then tested, but it was determined that there was not

enough confidence in the results to include it in the final design.

Post-print copy


Figure 11 - Energy intensity of HPGR, crush and SAG based circuits (SABC) from the feasibility study options overlaid onto the Tonne Intensity Energy Curve (Ballantyne et al., 2016).

A survey was conducted at the Tropicana mine to compare the performance of the HPGR and ball mills

with the design. The ore competence was similar to the design, the Bond Standard Work index was

18.8 kWh/t and A*b was 39.8, compared with the design values of 18.9 kWh/t and 31.5 respectively.

Laboratory locked-cycle HPGR tests were also conducted on the ore. The energy required to produce

new minus 75 µm material, or Size Specific Energy (SSE75), was used to compare the laboratory,

design and operating efficiency of the circuit (see Figure 12). The efficiency of the HPGR during the

survey was similar to both the overall circuit design and the HPGR laboratory results, reflecting the

accuracy that is possible for design and operation of HPGR circuits. This is made more remarkable

since the throughput and grind sizes are significantly different between the different cases. There are

two results for the ball mill that reflect the performance of the ball mill in two separate surveys, one

conducted with the ball mill operating at 83% critical speed (Nc) and the second at 75% critical. The

performance of the mill improved at the lower speed, with a higher throughput, finer grind at lower

power consumption achieved at 75% critical speed. This result was thought to be due to a

combination of poor motor power efficiencies and reduced grinding performance. A slip energy

recovery (SER) drive is used to control the mill speed by running at slower than synchronous speeds,

therefore rotor current and shaft power are not directly proportional. The site have also identified

differences in ball size, trajectory and liners profiles at this point in time. It is clear from these results

that an operating ball mill can perform more, or less, efficiently than the HPGR in response to changes

in internal grinding conditions, independently from HPGR operation.

HP

GR

Cru

sh

SAB

C

Post-print copy


Figure 12 - Comparison between circuit design, HPGR laboratory locked-cycled tests and survey results from Tropicana mine on the SSE75 Intensity Energy Curve. Two surveys were completed with the ball mills operating at different critical speeds (Nc).

Conclusions The high efficiency of the grinding mechanism of the HPGR has been demonstrated in previous studies

and is not disputed, but the ball milling requirements following HPGR has not hitherto been clearly

outlined. Design studies investigating the trade-off between SAG and HPGR based circuits are regularly

being conducted with lower Bond work index in the HPGR circuits. The reduction in BBWi is reportedly

explained by the presence of microcracks generated under the high forces of HPGR crushing, however

the increased fines in the HPGR product, relative to stage crushing, also plays a large role in reducing

the apparent BBWi. The standard BBWi for stage-crushed feed is used for SAG circuits because of the

apparent lack of microcracks. The high proportion of surface breakage that occurs in SAG milling can

result in the generation of more fines than the HPGR product. Thus if the difference in BBWi is solely

due to the difference in fines in the feed and not a change in competence, the BBWi of SAG product

could be less than HPGR product, not greater. Therefore, the effect of fines in the feed needs to be

isolated from the effect of microcracks when using the results from the BBWi to determine the ball

milling requirements following HPGR.

The hypothesis that this paper aimed to explore is that using per cent passing a marker size such as

75 µm is a more effective way of assessing the grindability of an ore than a P80 because it is more

closely related to the surface area generation. The feed to Bond Ball Mill tests had similar P80s even

when the amount of fines were different. When the natural size distribution from staged-crushing

and HPGR were characterised, the reduction in BBWi for the HPGR was much greater than the

reduction in SSE75. When the HPGR product was reconstituted to have the same size distribution as

the crushing product, similar degrees of reduction were observed for BBWi and SSE75. Thus, using

Bal

l mill

(7

5% N

c)

Bal

l mill

(8

3% N

c)

HP

GR

Post-print copy


the SSE75 is preferred to the BBWi to characterise the grinding requirements of HPGR product versus

crusher product. A methodology to calculate the SSE75 from a standard Bond Ball Mill test has been

described. This methodology will help identify undersized ball mills following equipment that

generate different size distribution gradients from the standard staged crushing product.

This analysis exclusively focused on the Bond ball test, but in so doing, highlighted many flaws in this

test for the characterisation of HPGR products. The HPGR action is likely to result in weaker coarser

progeny, due to microcracking, and intact fines. In addition, the HPGR product is typically screened

at coarser sizes than 5 mm. The fact that the standard Bond test is conducted on -3.35 mm material

means that the selective weakening of coarser particles is not isolated. The mineral processing

practitioner should make every effort to separate the effect of changes in the shape of the size

distribution from the strength of the rock when characterising ore for ball milling following HPGR.

More measurement of the degree of weakening and microcracking would be obtained by combining

single particle strength measurements with bulk tomography measurements. This would enable the

strength by size relationship to be determined and correlated to a measure of degree of micro-

cracking. This methodology would allow the full HPGR progeny to be measured and could also be

used to measure the degree of preferential liberation that occurred. These results should then be

compared to the product from a pilot scale SAG mill to reduce the bias that could be created from

scalping the crushed feed.

References Amelunxen, P., Meadows, D., 2011. Not another HPGR trade-off study! Minerals & Metallurgical

Processing 28, 1-7.

Ballantyne, G.R., Peukert, W., Powell, M.S., 2015. Size specific energy (SSE)—Energy required to

generate minus 75 micron material. International Journal of Mineral Processing 139, 2-6.

Ballantyne, G.R., Powell, M.S., 2014. Benchmarking comminution energy consumption for the

processing of copper and gold ores. Minerals Engineering 65, 109-114.

Ballantyne, G.R., Powell, M.S., Clarke, N., Di Trento, M., Kock, F., Putland, B., 2016. Energy curve case

study (AngloGold Ashanti), SME Annual Conference and Expo (ACE), Phoenix, Arizona.

Bartholomew, K.M., McIvor, R.E., Arafat, O., 2014. Functional Performance of Ball Milling Circuits - A

plant metallurgist's tool for process characterisation and optimisation, Mill Operators Conference,

Townsville, Australia, pp. 227-231.

Battersby, M.J.G., Kellerwessel, H., Oberheuser, G., 1992. Important Advances in the Development of

High Pressure Rolls Comminution for the Minerals Industry, Extractive Metallurgy of Gold and Base

Metals, Kalgoorlie, Australia, pp. 159 - 166.

Baum, W., and Ausburn, K., 2011. HPGR Comminution for Optimization of Copper Leaching, Minerals

and Metallurgical Processing 28(2):77-81

Post-print copy


Baum, W., Patzelt, N., Knecht, J., 1997. Metallurgical benefits of high pressure roll grinding for gold

and copper recovery, Society of Mining Engineering (SME), Denver, Colorado, pp. 111-116.

Bond, F.C., 1949. Standard Grindability Tests Tabulated, AIME Transactions, p. 313.

Bond, F.C., 1952. The third theory of comminution. Transactions AIME, 484-494.

Bond, F.C., 1961. Crushing and Grinding Calculations Parts 1 and 2. British Chemical Engineering 6,

378-385, 543-548.

Bueno, M.P., Kojovic, T., Powell, M.S., Shi, F., 2013. Multi-component AG/SAG mill model. Minerals

Engineering 43–44, 12-21.

Clarke, A.J., Wills, B.A., 1989. Technical note Enhancement of cassiterite liberation by high pressure

roller comminution. Minerals Engineering 2, 259-262.

Daniel, M.J., 2007. Energy efficient mineral liberation using HPGR technology, Julius Kruttschnitt

Mineral Research Centre. The University of Queensland, (Unpublished Thesis).

Davaanyam, Z., Klein, B., Nadolski, S., 2015. Using piston press tests for determining optimal energy

input for an hpgr operation, SAG conference, Vancouver, Canada.

Esna-Ashari, M., Kellerwessel, H., 1988. Interparticle crushing of gold ore improves leaching, Randol

Gold Forum Scottsdale.

Garcia, D., Lin, C.L., Miller, J.D., 2009. Quantitative analysis of grain boundary fracture in the breakage

of single multiphase particles using X-ray microtomography procedures. Minerals Engineering 22, 236-

243.

Hogg, R., Fuerstenau, D.W., 1972. Power relationships for tumbling mills. SME-AIME 252, 418-432.

Kodali, P., Dhawan, N., Depci, T, Lin, C.L., and Miller, J.D. Particle damage and exposure analysis in

HPGR crushing of selected copper ores for column leaching, Minerals Engineering, Volume 24, Issue

13, October 2011, Pages 1478-1487

Koski, S., Vanderbeek, J., Enriquez, J., 2011. Cerro Verde concentrator - four years operating HPGRs,

SAG 2011. Department of Mining Engineering, University of British Columbia, Vancouver, B. C.,

Canada.

Levin, J., 1989. Observation on the Bond standard grindability test, and a proposal for a standard

grindability test for fine materials. SAIMM 89, 13-21.

Levin, J., 1992. Indicators of grindability and grinding efficiency. Jounal of South African Institute of

Mining and Metallurgy 92, 283-290.

Lin, C.L., Miller, J.D., Hsieh, C.H., 2012. Particle damage during HPGR breakage as described by specific

surface area distribution of cracks in the crushed products, IMPC 2012, New Delhi, India, pp. 3397 -

3410.

Post-print copy


Morrell, S., 1992. Prediction of grinding-mill power. The Institution of Mining and Mettalurgy (Section

C: Mineral processing and extractive metallurgy) 101, C25-C32.

Morrell, S., 2009. Predicting the overall specific energy requirement of crushing, high pressure grinding

roll and tumbling mill circuits. Minerals Engineering 22, 544-549.

Musa, F., Morrison, R., 2009. A more sustainable approach to assessing comminution efficiency.

Minerals Engineering 22, 593-601.

Oestreicher, C., Spollen, C.F., 2006. HPGR versus SAG mill selection for the Los Bronces grinding circuit

expansion, SAG 2006. Department of Mining Engineering, University of British Columbia, Vancouver,

B. C., Canada, pp. 110-123.

Otte, O., 1988. Polycom high pressure grinding principles and industrial application, Third Mill

Operators' Conference. AusIMM, Cobar, NSW, pp. 131-136.

Palm, N., Shackleton, N., Malysiak, V., O’Connor, C.T., 2010a. The effect of using different

comminution procedures on the flotation of platinum group minerals. In: XXVth International Mineral

Processing Congress. AusIMM, Brisbane, Australia, pp. 1077–1083.

Patzelt, N., Klymowsky, R.I.B., Knecht, J., Burchardt, E., 2005. HPGRs for hard rock applications, Randol

Innovative Metallurgy Forum, Perth, pp. 80-89.

Patzelt, N., Klymowsky, R.I.B., Knecht, J., Burchardt, E., 2006. High-pressure grinding rolls for

gold/copper applications, in: Kawatra, S.K. (Ed.), Advances in Comminution, pp. 51-68.

Patzelt, N., Knecht, J., 1996. The use of high pressure roll grinding for optimization of copper leaching

in: Baum, w. (Ed.), SME Annual Meeting, Phoenix, Arizona.

Rajamani, K., Herbst, J.A., 1984. Simultaneous estimation of selection and breakage functions from

batch and continuous grinding data. Transactions of the Institution of Mining and Metallurgy, Section

C: Mineral Processing and Extractive Metallurgy 93, 74-85.

Rowland, C.A., Kjos, D.M., 1980. Rod and ball mills, in: Mular, A.L., Bhuppa, R.B. (Eds.), Mineral

Processing Plant Design, AIME, pp. 239-278.

Schönert, K., 1988. A first survey of grinding with high-compression roller mills. International Journal

of Mineral Processing 22, 401-412.

Shi, F., Lambert, S., Daniel, M., 2006. A study of the effects of hpgr treating platinum ores, SAG 2006.

University of British Columbia, Vouncouver, Canada.

Solomon, N., Becker, M., Mainza, A., Petersen, J. and Franzidis, J.-P. 2011. Understanding the influence

of HPGR on PGM flotation behavior using mineralogy, Minerals Engineering, 24. pp 1370-1377.

Stephenson, I., 1997. The Downstream effects of high pressure grinding rolls processing (PhD thesis),

Julius Kruttschnitt Mineral Research Centre, Department of Mining, Materials and Minerals

Engineering. University of Queensland.

Post-print copy


Tavares, L.M., 2005. Particle weakening in high-pressure roll grinding. Minerals Engineering 18, 651-

657.

Tüzün, M.A., 2001. Wet Bond mill test. Minerals Engineering 14, 369-373.

van der Meer, F.P., Gruendken, A., 2010. Flowsheet considerations for optimal use of high pressure

grinding rolls. Minerals Engineering 23, 663-669.

van der Meer, F.P., Schnabel, H.G., 1997. The effect of roller press grinding on ball milling energy.

Erzmetall 50, 554-561.

Vanderbeek, J.L., Linde, T.B., Brack, W.S., Marsden, J.O., 2006. HPGR implementation at Cerro Verde,

SAG 2006. Department of Mining Engineering, University of British Columbia, Vancouver, B. C.,

Canada, pp. 45-61.

Vizcarra, T.G., Wightman, E.M., Johnson, N.W., Manlapig, E.V., 2010. The effect of breakage

mechanism on the mineral liberation properties of sulphide ores. Minerals Engineering 23, 374-382.

Wang, C., Nadolski, S., Mejia, O., Drozdiak, J., Klein, B., 2013. Energy and cost comparisons of HPGR

based circuits with the SABC circuit installed at the Huckleberry mine, 45th Annual Canadian Mineral

Processors Operators Conference, Ottawa, Ontario, pp. 121-135.

Watson, S., Brooks, M., 1994. KCGM Evaluation of High Pressure Grinding Roll Technology, Fifth Mill

Operators' Conference. AusIMM, Roxby Downs SA, pp. 69-83.

Weier, M., 2016. JKTech 2015/16 Comparative Testing Program Report, in: Chenje, T., Napier-Munn,

T., Morrell, S. (Eds.). JKTech Job 14294.

Xu, W., Dhawan, N., Lin, C.-L., Miller, J.D., 2013. Further study of grain boundary fracture in the

breakage of single multiphase particles using X-ray microtomography procedures. Minerals

Engineering 46–47, 89-94.

Acknowledgements With thanks to the Coalition for Energy Efficient Comminution (CEEC) who provided the resources to

complete this analysis and Weir Minerals, AngloGold Ashanti and Orway Mineral Consultants (OMC)

who showed enthusiasm and openness to collaborate in this investigation. As stated in the authorship

notes: comparative Bond work index data were supplied by Frank van der Meer, Evert Lessing and Ric

Stocco of Weir Minerals. And design, laboratory and operating data from Tropicana Mine were

supplied by Nick Clarke, Mike Di Trento of AngloGold Ashanti, and Fred Kock and Brian Putland of

Orway Mineral Consultants (OMC).

Improved characterisation of ball milling energy ...623483/UQ623483_OA.pdf · This paper describes...

Documents

Transcript of Improved characterisation of ball milling energy ...623483/UQ623483_OA.pdf · This paper describes...