Minerals Engineering 119 (2018) 228–235

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

The comminution energy-size reduction of the Bond Mill and its relation to T
Vickers Hardness
⁎
Mario Menéndeza, Héctor Muñiz Sierrab, Malcolm Genta, , Francisco Javier de Cos Jueza
a
School of Mining, Energy and Materials Engineering, Universidad de Oviedo, c/Independencia 13, Oviedo 33004, Spain
b
School of Mines, Energy and Materials, University of Oviedo, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: The results of Bond Work index (BWi) tests of five materials with five closing screen size (P1) of 63o 500 µm are
Comminution energy-size relation documented. These are normalized to eliminate the effect of material present finer than the closing screen sizes
Bond Mill specific grinding energy and test sample apparent densities. Analysis of normalized test results show, with a correlation factor R2 of
Vickers Hardness 0.9956, that the revolutions required in all the Bond Work index mill tests conducted (29–419 revolutions)
corresponds to a specific grinding energy per Bond Mill revolution of 0.0823 kWh/ton for a material corrected to
a tapped apparent density of 1 t/m3 and for material present finer than the closing screen size. Conversion of this
value to operation of the mill for the Bond Work index test results in a value of 194 ∗ 10−7 kWh/mill revolution.
It is demonstrated that the indicated specific grinding energy of the Bond Mill revolution is not a precise
value. An exponential function with a perfect correlation is presented relating the final required number of Bond
Work index test mill revolutions required with the P1 values for sample materials tested.
A perfect correlation with an exponential relation to the evolution of the specific grinding energy with change
in closing screen size is presented. This relation is interpreted as a measure of the reduction in grinding efficiency
with increase in fineness of grinding.
It also shown that there is an exponential relation with a R2 correlation factor of 0.9900 between experi-
mental specific grinding energy of a mono-mineralogical material normalized for < P1 content and apparent
density to the root of its VH at a given P1. Most if not all the deviations between the calculated specific grinding
energy based on P1 values and the sample Vickers Hardness correlates with the differences between the ex-
perimental and corrected mill revolutions applied to compensate for the < P1 fines contents of the test samples.

1. Introduction
Grinding is one of the largest industrial consumers of energy. Some
4% of the total world’s electrical energy consumption is consumed in
the crushing and grinding of mineral ores (Energy Efficiency Exchange,
2013). Large amounts of energy are also consumed in the production of
many commercial products such as cements, fillers, pigments, fritzes,
cereal flours and pharmaceuticals. Being energetically very inefficient,
the largest portion of this energy consumed is typically in grinding.
Effectively accounting for and predicting the energy necessary to
fragment a material to a given particle size has been the basis of ex-
tensive research by engineers, chemists, physicists and mathematicians
into the theoretical and practical aspects of comminution and especially
that of energy consumption in particle size reduction by grinding.
To date the theories of Kick (1885), von Rittinger (1987) and Bond
(1952) remain the most established. The recommendations of Charles
(1957) and Hukki (1962, 1975) that Kicks’s is the most applicable for
crushing, Bond’s for ball mill grinding and von Rittinger’s for fine
grinding is generally accepted. Of these theories, that of von Rittinger
has a theoretical basis founded on the new particle surface area gen-
erated whereas that of Kick’s is based on the change in volume of
particles created. That of Bond is empirical and based on very extensive
laboratory test results compared to industrial grinding operations of the
same materials tested in the laboratory. However, neither Kick’s nor
von Rittinger’s theories include an established means of determining
the energy requirements for grinding on an industrial scale. Only Bond’s
theory is widely used for effectively designing or controlling ball mill
grinding circuits despite it not including a theoretical basis and un-
derestimating energy consumption for production of particles finer than
100 µm (Hukki, 1962, 1975).
The procedure to apply Bond’s law requires the laboratory de-
termination of a Bond Work index (BWi). This laboratory procedure has
been extensively described in detail by Bond (1952), Deister (1987) and
Mosher and Tague (2001). It is conducted in a specified ball mill and

⁎
Corresponding author.
E-mail address: gentmalcolm@uniovi.es (M. Gent).

https://doi.org/10.1016/j.mineng.2018.01.017
Received 24 May 2017; Received in revised form 18 December 2017; Accepted 16 January 2018
0892-6875/ © 2018 Elsevier Ltd. All rights reserved.
M. Menéndez et al.
Fig. 1. Examples of the variation in BWi values with closing screen size (P1).
grinding ball charge. The BWi is calculated using Eq. (1) based on: the
cumulative 80 percentile grain undersize (F80) expressed in µm of a
minus 3.36 mm mill feed; the test product closing screen cumulative 80
percentile grain undersize (P80) expressed in microns; and the grams
per revolution (G) produced finer than the closing screen size (P1). The
test is conducted using a locked cycle with a 250% recirculating
700 cc.mill feed until the values of G produced stabilize. Bond tests at
P1 ≤100 µm are common but can require wet screening of each test
cycle product and frequently require more than 10 days to complete.
BWi(kWh/tonne) = 49.1/[P10.23 xG 0.82∗ ((10/P80
0.5 0.5
)−(10/F 80 ))]
Ball mills are typically used to grind materials to < 500 µm. Hukki
proposed limiting Bond’s law to grinding no finer than 100 µm and BWi
values tend to be relatively constant for grinding products to between
500 and 200 µm. However, experience has shown that some materials
(Fig. 1) have a significant increase in BWi values between 250 and
125 µm and increase exponentially for lower values.
The BWi test is a lengthy process and a certain degree of variation in
results always occurs due to equipment wear as well as operator and
operational errors or variations between operator and operational
procedures. Results for a given sample may vary due to errors or fluc-
tuations in: sampling; particle size analysis; determination of the test
sample weight; inefficiency of screening of the mill product (especially
for grinding < 250 µm); losses of mill product fines and recirculated
load; effects of mill charge and liner wear; mill liner roughness; etce-
tera. The test grinding efficiency is also affected by variations in mill
charge weight and thus the total grinding surface area available.
Variations in degree of sample tapping to determine the weight of
sample to be used in the BWi test has a direct proportional influence on
the grams per revolution obtained. The weights of mill product fines
produced for abrasive materials are also influenced as a result of con-
tamination by iron from the mill charge and liner.
To date no rapid method has been developed to predict the en-
ergy–size reduction relation of a given material without laboratory
testing. A number of different alternative theories of grinding energy to
particle size relations have been proposed. Despite the complexity of
Bond’s test procedure, most of these are based on the use of his specified
test mill and its feed specifications. Simplified or alternative methods
proposed to determine a comminution work index have been proposed
by a number of investigators. Examples of the diverse nature of some of
the recent investigations are those of Ahmadi and Shahsavari (2009),
Aksani and Sönmez (2000), Deniz and Ozdag (2003), Gharehgheshlagh
(2016), Morrell (2004), Namura and Tanaka (2011), Ozkahraman
(2005, 2010), and Zambrano (2002). A computer simulation model
simulating the Bond grindability test based on cumulative kinetics in-
corporating parameters derived from batch grinding tests has been
proposed by Aksani and Sonmez. Zambrano developed a general power
model that predicts energy input for grinding based on a statistical
Minerals Engineering 119 (2018) 228–235
analysis of energy–size reduction relationships in batch tumbling ball
mills. He reported an error < 5% for von Rittinger’s law, 15–20% error
for Bond’s law and 20–30% for Kick’s law. He also reported that a lab
mill behaves like an industrial-size mill. An alternative method of de-
termining the Bond grindability and BWi was developed by Deniz and
Ozdag based on material dynamic elastic parameters. Morrell proposed
an alternative energy-size relationship for 100 to +0.1 mm particles to
that of Bond’s based on a particle size exponent that is a function of feed
and product particle sizes. Unfortunately, the author did not present a
value(s) for the function proposed. Nor is the proposal demonstrated to
be applicable to grinding to < 100 µm. Ozkahraman showed that the
BWi and the grindability index could be estimated from a materials
friability value. Ahmadi and Shahsavari used first-order grinding ki-
netics based on results of the two Bond grinding cycle tests to develop a
rapid method of determining the BWi. Namura and Tanaka presented
an empirical power law between energy input and size reduction. The
power applied varies according to the material ground and requires
knowledge of its strength properties. Gharehgheshlagh found that a
kinetic batch grinding test for 5 different grinding internals could be
used to calculate the BWi with < 2.60% error. He also observed that the
mill product d80 is a function of P1.
This investigation reports the results of BWi tests of five materials of
diverse physical characteristics conducted with the objective of de-
termining if the use of corrections for variations in sample character-
istics could be applied to determine the equivalent energy consumed
per Bond Mill revolution. Vickers hardness tests of the samples were
also conducted to determine the extent to which the grams of fines
(1) produced at a given closing screen size as a function of the materials
hardness values. The determination of the specific grinding energy of
the Bond Mill in tests where a complete Bond Work index test may not
be practical has potential applications in some instance. This is espe-
cially so in the cases of tests where fineness of the grinding to be
achieved does not permit effective preparation of recirculated material,
the presence of materials extremely resistant to grinding such as metals
in slags, time requirements, open circuit grinding, etcetera.
Comparative grinding testing methods such as proposed by Berry
(1966) to determine the energy consumption for fine materials such as
plant middlings and tailings, as well as most commercial products re-
quire the use of a reference material with a similar particle size dis-
tribution and grinding characteristics. Reference samples with a sui-
table particle size distribution may not be available. Furthermore,
reference samples of ores composed of minerals with large variations in
resistance to grinding can be especially difficult to identify.
Magdalinović et al. (2011) observed that there might be a material
constant which characterises mineral resources grindability. Gent et al.
(2012a,b) observed from their BWi tests in addition to published data
that there appeared to be a significant polynomial correlation between
BWi and Vickers Hardness (VH) values. This relation was interpreted to
be due to VH values and BWi values being related to the plasticity of
materials and thus their fracture toughness but did not consider whe-
ther or not the VH of a material influenced the increase in BWi values
with increase in the degree of particle size reduction.
2. Theory
Grinding energy requirements for a given material decrease with
abundance of fissures, stress planes or inclusions. As breakage tends to
occur at the interface between crystals when particle sizes are greater
than the size of crystals composing them and the adherent energy is
usually lower than the coherent energy required when the crystal
themselves are broken, less energy is required to separate crystals than
to break them (Hukki, 1962 and Ozkahraman, 2010). The maximum
energy requirement for a particle size reduction occurs when the par-
ticles to be fragmented consist of liberated grains (typically individual
crystals). It is to be anticipated that BWi values should increase as the
degree of grinding fineness and proportion of liberated grain particles
229
M. Menéndez et al.
increases and that the BWi value for totally liberated particles should
represent the maximum value for a given P1 value.
Attempts to determine a unifying relation between energy con-
sumption and particle size reduction for ball mill grinding to any par-
ticle size has had limited results. It is proposed that this has been due in
part to the fact that many investigations were based on the analysis of
BWi test results did not generally included corrections for physical
differences in samples tested. Levin (1989) proposed that since there is
an established relation between Bond test mill results and industrial
scale operations, there should be an equivalent energy consumed per
revolution of the Bond Mill which could be applied without the use of a
reference sample. He proposed that the corrected BWi test G values in
the raw mill feed for the presence of material < P1 should be increased
by the total amount of < P1 material in the test sample. However, this
correction supposes an unrealistic solution that the total amount of U
added to G is ground to product in fines. His calculations based on 248
BWi tests of a wide variety of materials resulted in a range of values
with most clustering around 198 × 10−7 kWh/Bond Mill revolution.
No correction for the change in P80 values of the raw feed were made
and would in any case have a negligible to no appreciable effect.
However, it is argued that the actual U entering the mill and which
cannot contribute to fines produced when the locked cycle grinding
stabilizes consists of the proportion of new sample material that ac-
companies the recirculating fraction (i.e. 1/3.5 of that in the original
sample). A corrected G value for U content (Gf) should be in proportion
to the percentile of these fines within the new mill feed as expressed in
Eq. (2).
Gf = G/(1−(0.01 × U/3.5))
As the BWi test is based upon the use of a fixed volume of 700 cc of test
material but the results are measured in weight, two sample materials
of the same physical grinding characteristics but having different test
sample weights (i.e. different tapped apparent densities) must have
different G values. To correct BWi test Gf values for differences in ap-
parent densities (Gγ), Gf values should be corrected in proportion to the
original sample apparent densities (γ) determined from the initial
700 cc sample weight and as expressed by Eq. (3).
Gγ = Gf / γ = G/ γ [1−(0.01 × U/3.5)]
The product of the number of test mill revolutions (Rev) and the cor-
rected value of G with Eq. (2) does not correspond to a 250% re-
circulating load. For a BWi test sample of 700 cc normalized to an ap-
parent density of 1 g/cm3, based on a 250% recirculating load, the
corrected Rev value (Revc) should be the quotient of 200 divided by Gγ,
200 being the product of closed circuit for a material with corrected
apparent density to 1 g/cm3. It is emphasized that this only an ap-
proximation with the potential for errors increasing with the content of
U in the sample. In addition, it is to be anticipated that this projected
increase in particle size of the mill feed should result in the necessity of
a greater Rev to complete the BWi test.
The Bond Work index (BWic) corrected for U and γ can be expressed
as in Eq. (4).
BWi c = 49.1/[P10.23 × {G/ γ [1−(0.01 × U/3.5)]}0.82 × ((10/P80
0.5 0.5
)−(10/F 80 ))]
Magdalinović et al. (2011) has shown that for three materials tested at
three different P1 values, there appear to be linear correlations between
the G and square root of the test P1 value. This relation between G and
P1 can also be observed for other sample materials tested at different P1
sizes such as some of those reported by Levin (1989). This linear rela-
tion suggests that:
• Either the effects due to the presence of fissures, stress planes, inter-
grain bonding does not significantly affect the G produced, or
• The rocks tested were not ground to a fineness at which these de-
fects were absent or diminish in abundance.
Minerals Engineering 119 (2018) 228–235
The differences in correlations observed between the three materials
tested in his study could be interpreted as to being due to differences in:
• Abundances in fissures, stress planes, inter-grain bonding, and/or
• Grain size distribution of minerals present in samples, and/or
• Physical characteristics of the grains constituting the samples tested.
It is interpreted that these differences in correlations of different
materials might be attributed to variations between their physical
characteristics, of which it was anticipated that the materials elasticity
as indicated by its hardness is a potential factor.
3. Methodology
To negate effects in BWi tests due to fissures, stress planes and grain
bonding effects in samples, five samples locally available were selected
as being relatively free of these mechanical weaknesses. They consisted
of large, individual crystals of calcite, labradorite, magnesite and
quartz. These provided a range of compositions, differences in crystal-
lographic cleavages, and hardness’s. A sample of plate glass was also
selected as being an example of a homogenous material free of inclu-
sions.
The samples were prepared by an initial washing to remove any
adhering particles, dried and then crushed to < 25 mm. They were then
hand sorted to remove any contaminants, and crushed to meet Bond’s
mill feed specifications. The samples were subsequently divided into
representative fractions using a Jones riffle splitter with 10 mm chan-
nels to minimize any sampling error. Tests were conducted as per the
(2)
procedure described by Bond (1952, 1961) with 500, 250, 125, 90 and
63 µm closing screens (P1). Sample feed and mill product sieving was
conducted as per the ASTM E276-03 procedure. All the 90 and 63 µm P1
BWi test products were wet screened to ensure accurate determinations
of product weights and grain size distributions.
Blanks for the determination of the sample materials Vickers hard-
ness were prepared with 2.5 to 3.15 mm particles mounted in epoxy by
casting in 30 mm wide, 30 mm thick, plastic cups. To prepare the
samples for polishing the castings were mounted in a Buehler Isomet
(3) low speed diamond and saw cut into 20 mm thick disks with the top
parallel to the bottom. This procedure ensured that no sample heating
occurred and that the test surface had a minimum of irregularities. The
cut surfaces were polished using progressively fine grits finishing with a
0.1 µm diamond grinding media. The VH values of the sample materials
were determined from the average of 50 measurements per test mate-
rial according to the ASTM E384 procedure.
4. Results and discussion
The physical characteristics determined for the test sample mate-
rials based on their sieve analysis (Fig. 2), preparation of the BWi test
mill charge and determination of the Vickers Hardness are shown in
Table 1.
Initial BWi test results and calculated values are presented in
Table 2.
The grinding actions within a ball mill consist of crushing due to the
(4)
impact of the mill charge balls on sample grains and their shearing due
to their attrition by the mill balls and other mineral particles. The VH of
these samples is indicative of their elasticity and as such, resistance to
crushing, chipping and abrasion while the presence of cleavage planes
represents planes of weakness along which shearing or chipping and
abrasion may be especially effective. As such, it was anticipated that the
quartz and glass samples should have higher resistances to crushing and
shearing, magnesite and calcite samples should have lower resistances
to these two forces and that the labradorite sample intermediate char-
acteristics.
The results of the corrections of G to Gγ values with Eq. (3), BWi to
BWic values with Eq. (4) and Rev to Revc values are indicated in
230
M. Menéndez et al. Minerals Engineering 119 (2018) 228–235

Table 3
Calculated Gγ, BWic and Revc values derived from initial BWi test values corrected for
sample U and γ properties as well as the calculated Wc based on sample test F80 and P1
values by closing screen size (P1).

Material P1 (µm) Gγ (g/revolution) BWic (kWh/t) Wc (kWh/t) Revc

Quartz 500 2.50 18.32 5.55 80

Quartz 250 1.57 18.56 9.53 127
Quartz 125 0.90 21.18 17.62 222
Quartz 90 0.64 25.48 25.15 312
Quartz 63 0.46 30.16 35.91 437
Glass 500 2.56 17.83 5.43 78
Glass 250 1.55 19.00 9.60 129
Glass 125 0.84 22.41 18.75 240
Glass 90 0.65 25.20 24.93 309
Glass 63 0.47 29.69 35.33 428
Labradorite 500 2.38 19.29 5.78 84
Labradorite 250 1.51 19.62 9.83 132
Fig. 2. Particle size distributions of the test sample materials.
Labradorite 125 0.83 22.53 18.88 242
Labradorite 90 0.66 25.03 24.64 305
Labradorite 63 0.49 28.16 34.08 410
Table 1
Magnesite 500 3.61 14.34 4.10 55
Sample material characteristics.
Magnesite 250 2.00 15.25 7.80 100
Material Apparent density F80 (µm) Vickers Cleavage axis and Magnesite 125 1.24 16.49 13.57 162
(g/cm3) Harness (GPa) quality Magnesite 90 0.95 18.32 18.14 210
Magnesite 63 0.70 20.98 25.41 286
Quartz 1.52 2673 10.7 1-Indistinct Calcite 500 5.54 10.21 2.89 36
Glass 1.42 2602 6.2 None Calcite 250 3.08 10.78 5.49 65
Labradorite 1.61 2635 7.9 1-Perfect, 1-Good, Calcite 125 1.63 12.76 10.82 123
1-Distinct Calcite 90 1.30 14.43 14.05 154
Magnesite 1.88 2302 3.9 3-Perfect Calcite 63 0.97 16.30 19.38 206
Calcite 1.72 2250 1.8 3-Perfect

function with a R2 correlation factor of 0.9956, indicating a value of

Table 3. Calculated BWic values and corrected Bond’s required grinding
power (Wc) based on the sample test F80 and P1 values are also in-
dicated in this table.
4.1. Analysis of Bond Mill revolutions to specific mill grinding energy
The relation between the specific grinding energy (Wc) based on the
normalized BWic, to the corrected required BWi test revolutions (Revc)
is shown in Fig. 3. Assuming that no energy is consumed if there are no
mill revolutions, this relation can be expressed in Eq. (5) as a linear
0.0823 kWh/t revolution for a sample with no fines (U) and an apparent
density of 1 g/cm3. The spread in data relative to the projected mean
value for the specific grinding energy (B) indicates that other factors not
adequately included in Bond’s theory affects the specific grinding en-
ergy. Even for coarse grinding where the closing screen is 500 µm, there
is an evident but small diversion in the observed trends for each of the
sample materials tested as indicated in Fig. 4. This figure indicates that
the greatest divergence occurs with the softer samples (calcite and
magnesite) as compared to the tighter clustering of values for the
harder sample materials.

Table 2
Sample BWi test results and calculated BWi and Bond’s required grinding power W values based on the sample test F80 and P1 values.

Material U (%) P1 (µm) P80 (µm) Rev G (g/revolution) BWi (kWh/t) W (kWh/t)

Quartz 11.57 500 406 73 3.67 13.35 4.05

Quartz 6.17 250 200 121 2.35 13.35 6.85
Quartz 3.41 125 95 208 1.36 15.14 12.59
Quartz 2.61 90 72 292 0.97 18.17 17.93
Quartz 1.82 63 52 414 0.69 21.47 25.57
Glass 18.21 500 399 65 3.45 13.97 4.26
Glass 9.75 250 203 119 2.15 14.58 7.37
Glass 5.83 125 94 228 1.17 17.04 14.25
Glass 3.84 90 71 297 0.91 19.07 18.87
Glass 2.55 63 52 419 0.66 22.39 26.65
Labradorite 17.77 500 409 74 3.65 13.60 4.07
Labradorite 10.03 250 207 122 2.37 13.57 6.80
Labradorite 5.42 125 94 231 1.32 15.41 12.91
Labradorite 4.18 90 72 291 1.05 17.07 16.80
Labradorite 2.74 63 51 394 0.78 19.14 23.16
Magnesite 25.39 500 409 41 6.14 9.29 2.65
Magnesite 16.36 250 193 90 3.50 9.65 4.94
Magnesite 10.52 125 94 139 2.20 10.29 8.47
Magnesite 9.30 90 70 200 1.70 11.40 11.29
Magnesite 6.82 63 50 256 1.26 12.98 15.72
Calcite 24.21 500 410 29 8.88 6.93 1.96
Calcite 14.41 250 193 58 5.08 7.14 3.64
Calcite 7.60 125 89 115 2.75 8.32 7.06
Calcite 6.59 90 71 137 2.20 9.38 9.14
Calcite 5.00 63 51 188 1.65 10.56 12.56

231
M. Menéndez et al. Minerals Engineering 119 (2018) 228–235

0.088944 kWh/t revolution is indicated.

The correction of the 0.088944 kWh/t revolution value for the
average bulk density of the samples tested (1.62 g/cm3) yields an
averaged value of consumed energy by the Bond Mill of
194 × 10−7 kWh/ mill revolution. This value coincides very closely to
the modal average value of 198.4 × 10−7 kWh/Bond Mill revolution
reported by Levin (1989). However, Levin reported a range of
138 × 10−7 to 491 × 10−7 kWh/Bond Mill revolution based on 248
BWi tests. It is suggested that part if not most of the range in his re-
ported values is due to his method of correcting for the g/rev produced
for the total < P1 content of the full mill load, including the circulating
load rather than the proportion of new < P1 feed content, and for not
correcting for the apparent densities of the test samples
The variation in results reported by Levin (1989) is interpreted to be
due to differences in the grindability of the sample materials and the
Fig. 3. Trend of relation of specific grinding energy per ton of material with no fines < P1
increase in inefficiencies of grinding with reduction in P1. It is suggested
(U) and an apparent density of 1 g/cm3 to corrected Bond Mill revolutions (Revc).
that one or both of the powers in the terms P10.23 and (P800.5) in Bond’s
Eq. (1) may not necessarily be the exact values corresponding to the
specific grinding energy of a given material at any P1.
Analysis of the data found that the variations of Wc with Revc for a
given P1 can be expressed with the Eq. (6) where the constant C of this
relation varies with P1 as indicated in Table 5. This equation fits all P1
values with a perfect correlation (R2) of 1.0. These variations for a
given P1 value are interpreted to be a function of the sample material
grindability. It also indicates that for any given material tested for a
given P1 should correspond to this relation.

Wc = C× Rev 0.82
c (6)

Fig. 4. Variations in relations of specific grinding energy per ton to corrected Bond Mill
revolutions (Revc) for a P1 of 500 µm of the test sample materials with no fines < P1 (U)
and an apparent density of 1 g/cm3.
B(kWh/t) = 0.0823 × corrected mill revolutions (5)
Correlations of Wc to final test Bond Mill revolutions (Revc) for the
sample data (Table 4) using linear equations both with and without
requiring that for no mill revolutions Wc = 0 (Linear relation 1 and 2),
and a power-law equation indicates that there is an almost linear re-
lation between Wc and Revc but that there is a slight progressive power-
law increase in Wc with increase in the required Revc. This corresponds
in general to a decrease in P1 and an increase in sample hardness. All
three equations show variations in values with the type of sample
material tested. These variations are interpreted to be a measure of the
differences in grindability of the sample materials tested, the increase in
grinding inefficiency as product fines content increases with mill re-
volutions and possible tests errors. The power-law equations presented
have the highest correlation factors of the three types presented and
probably represent the truest value for a given BWi test with a large or
very small number of mill revolutions. Its average value for the range of
mill revolutions applied in this investigation (29–419) coincides most
closely to that of the average value from Linear relation 2. Based on the
linear trend for all the test sample data of this investigation, a value of
The variation in the constant C of Eq. (6) with P1 values from Table 4
was found to correlate perfectly (R2 = 1.0) to Eq. (7). This relation is
interpreted to be the function of increase in inefficiency of grinding
with decrease in P1 value. This is interpreted to be due to the effect that
as the P1 value decreases, the particle size of the recirculating load is
reduced and there is a larger number of sample particles present in the
mill and consequently a decrease in the proportion of mill balls im-
pacting sample particles per mill revolution.
C= 0.637 × P1−0.23 (7)
The combination of Eqs. (6) and (7) relating Wc to P1 and required BWi
test mill revolutions is indicated by Eq. (8).
Wc = (0.637 × P1−0.23) × Rev 0.82
c (8)
4.2. Analysis of material VH on BWi test grams per revolution
As reported by Magdalinović et al. (2011) and shown in Fig. 5, Gγ
values present almost linear correlations to the root of the P1 experi-
mental values used. The values for softer minerals (calcite and mag-
nesite) present slightly greater deviations from a linear fit P1 at va-
lues < 125 µm. This deviation from a linear trend is not attributed to
these samples having strong cleavages as this effect is not as pro-
nounced for the labradorite sample which is also characterized by
strong cleavages. Application of a linear equation to the Gγ relative to
P1 values for the sample materials yields Eq. (9) where the different m
and c values and the correlation factor for these are listed in Table 6.

Table 4
Equations relating Wc to X (Revc) for the BWi test samples and their correlation factors (R2).

Sample Power-law relation R2 Linear relation 1 R2 Linear relation 2 R2

Quartz 0.0452X1.1007 0.9995 0.0808X 0.9966 0.0851X−1.3198 1.0000

Glass 0.0459X1.0979 0.9999 0.0808X 0.996 0.0854X−1.4003 0.9997
Labradorite 0.0412X1.1175 0.9998 0.081X 0.9939 0.0868X−1.7109 0.9995
Magnesite 0.0464X1.1155 0.9999 0.0865X 0.9937 0.0928X−1.287 0.9995
Calcite 0.0574X1.0919 0.9999 0.0917X 0.9952 0.0971X−0.7994 0.9991

232
M. Menéndez et al. Minerals Engineering 119 (2018) 228–235

Table 5
Variations in Eq. (7) with P1.

P1 (µm) Eq. (6) Correlation factor (R2)

500 Wc = 0.1526 × Revc0.82 1.0

250 Wc = 0.1789 × Revc0.82 1.0
125 Wc = 0.2099 × Revc0.82 1.0
90 Wc = 0.2263 × Revc0.82 1.0
63 Wc = 0.2457 × Revc0.82 1.0

Fig. 6. Relation between Gγ values and the result of P1 (µm) values divided by the root of
the test material VH (GPa).

4.3. Analysis of the application of material VH to its specific grinding energy

Eq. (13) permits a theoretical estimation of the grams produced per

revolution at a given P1 for a mill with a 250% recirculating load,
grinding a sample of crystals composed of one mineral or component
with a known VH. From this calculation, the number of mill revolutions
(RevVH) to produce the 200 g of material of an apparent density of 1 g/
Fig. 5. Gγ values for < P1 (µm) content and normalized to an apparent density of 1 g/cm3 cm3 required for the 250% recirculating load can be determined and
relative to the root of the closing screen size (P1). applied to Eq. (8) to determine an estimated grinding energy WVHγ for a
sample material with an apparent density of 1.0 g/cm3 at a given P1.
Table 6 This operation is expressed as Eq. (14).
Values of m, c and correlation factors for Eq. (9) corresponding to the test sample ma-
WVHγ = (0.637 × P1−0.23) × {[200/[0.0398 × (P1/VH0.5)0.8095]}0.82 (14)
terials.
The results of these calculations (Table 7) show correlations
m c R2
(R2) > 0.99 and average errors of < 0.3% between the estimated and
Quartz 0.1402 −0.6671 0.9996 experimental grams/revolution, mill revolutions and grinding energy.
Glass 0.1473 −0.7538 0.9976 The differences between the experimental Revc, Gγ and Wc values and
Labradorite 0.1326 −0.6007 0.9981
those derived from the VH values (RevVHγ, GVHγ and WVHγ) suggests
Magnesite 0.2001 −0.9752 0.9881
Calcite 0.3199 −1.7685 0.9900
that there is a certain degree of robustness to the equations applied.
However, all these show significant ranges in errors, especially for the
samples of quartz and glass. It is of note that compared to the other
Gγ = ( m× P10.5)−c (9) sample materials, both of these are characterized by not having clea-
vages and may be the least friable of those materials tested.
Comparison of the values of m and c of Eq. (9) relative to the VH (GPa)
of the sample found that is a R2 correlation of 0.9856 for m and 0.9735 5. Discussions
for c as expressed in Eqs. (10) and (11). Compiling Eqs. (9)-(11) yields
the expression Eq. (12). The application of Eq. (12) to the P1 and VH The variation between experimental specific grinding energy re-
data values was found to have a R2 correlation of 0.9948 to the test Gγ quired and calculated values (Fig. 7) indicates that these values tend to
values with an average difference 0.67% and a standard deviation be- increase with fineness of P1 values and the differences between types of
tween them of 6.8%. However there is a significant range of percentile materials also tend to disperse with finer P1 values. This trend for the
differences (−14.7 to 19.8%) between the Gγ and calculated values. WVHγ values to diverge from the experimental values with reduction in
P1 values might be attributed to differences in: sample mechanical
m= (0.3829 × 1/VH0.5) × 0.0041 (10) properties; increased probability of BWi test operator errors; and in-
creased contamination of mill ball wear with increased mill revolutions
c= (−2.3181 × 1/VH0.5) + 0.1603 (11) in the < P1 product. However, the very high correlation factor of
0.99997 between the percentile difference between that of Rev to Revc
and the percentile difference between WC and WVHγ suggests that im-
Gγ = [((0.3829 × 1/VH0.5) × 0.0041) × P10.5]−((−2.3181 × 1/VH0.5)
provements are required in calculating the adjustment to mill revolu-
+ 0.1603) (12) tions required after the correction for sample < P1 fines content.
The results indicated in this investigation are based on BWi tests of
It was found that a more simple relation between Gγ values as a func- samples consisting of only one component and composed of totally
tion of P1 values divided by the root of the test material VH (GPa) has liberated grains. The use of VH values for determining the specific
an exponential relation (Fig. 6) indicated with Eq. (13). This expression grinding energy of a material composed of more than one component
has a R2 correlation factor 0.9905. An average error of −1.27%, and a with distinct VH values requires investigations. Investigations such as
standard deviation of the percentile differences between Gγ values and those of Csoke et al. (2013) have shown that the BWi values of samples
those calculated with Eq. (13) (GVHγ) of 6.5%. The range of percentile composed of more than one material are not directly in proportion to
errors was found to be −12.1 to 12.9%. the amounts of these materials and their BWi values. It is suggested that
the use of VH values in the expression proposed by Tavares and
G VHγ = 0.0398 × (P1/VH0.5)0.8095 (13)
Kallemback (2013) for calculating the BWi value for blends based on a

233
M. Menéndez et al. Minerals Engineering 119 (2018) 228–235

Table 7
Comparison of calculated GVHγ, RevVHγ and WVHγ values and their errors relative to Gγ, Revc, and Wc values; and the correlation factors, standard deviations, means, maximum and
minimum differences corresponding to these.

Material P1 (µm) Revc Gg Wc GVHγ Error in GVHγ (%) RevVHγ Error in RevVHγ (%) WVHγ Error in WVHγ (kWh/t) Error in WVHγ (%)

Quartz 500 81 2.46 5.62 2.41 2.09 83 −2.14 5.72 0.10 −1.72
Quartz 250 127 1.57 9.53 1.37 12.45 145 −14.22 10.62 1.09 −11.49
Quartz 125 222 0.90 17.62 0.78 12.90 255 −14.81 19.73 2.11 −11.96
Quartz 90 312 0.64 25.15 0.60 6.08 333 −6.48 26.47 1.32 −5.26
Quartz 63 437 0.46 35.91 0.45 1.67 444 −1.70 36.41 0.49 −1.37
Glass 500 78 2.56 5.43 2.80 −9.30 71 8.51 5.05 −0.38 7.05
Glass 250 129 1.55 9.60 1.60 −2.88 125 2.80 9.38 −0.22 2.32
Glass 125 240 0.84 18.75 0.91 −9.27 219 8.48 17.43 −1.32 7.03
Glass 90 309 0.65 24.93 0.70 −8.13 286 7.52 23.38 −1.55 6.23
Glass 63 428 0.47 35.33 0.52 −12.13 382 10.82 32.16 −3.17 8.98
Labradorite 500 84 2.37 5.80 2.63 −11.14 76 10.02 5.31 −0.48 8.31
Labradorite 250 132 1.51 9.83 1.50 0.61 133 −0.62 9.88 0.05 −0.48
Labradorite 125 242 0.83 18.88 0.86 −3.52 233 3.40 18.35 −0.53 2.82
Labradorite 90 305 0.66 24.64 0.66 −0.12 304 0.12 24.61 −0.03 0.12
Labradorite 63 410 0.49 34.08 0.49 −0.81 406 0.81 33.85 −0.23 0.69
Magnesite 500 55 3.61 4.10 3.66 −1.13 55 1.12 4.06 −0.04 0.94
Magnesite 250 101 1.98 7.87 2.09 −5.30 96 5.04 7.55 −0.33 4.17
Magnesite 125 162 1.24 13.57 1.19 3.88 168 −4.04 14.02 0.44 −3.27
Magnesite 90 210 0.95 18.14 0.91 4.33 219 −4.53 18.80 0.67 −3.67
Magnesite 63 286 0.70 25.41 0.68 2.18 293 −2.23 25.86 0.46 −1.80
Calcite 500 36 5.54 2.89 5.15 6.93 39 −7.44 3.06 0.17 −6.04
Calcite 250 65 3.08 5.49 2.94 4.43 68 −4.63 5.69 0.21 −3.76
Calcite 125 123 1.63 10.82 1.68 −2.80 119 2.73 10.58 −0.25 2.26
Calcite 90 154 1.30 14.05 1.29 1.22 156 −1.23 14.19 0.14 −0.99
Calcite 63 206 0.97 19.38 0.96 0.87 208 −0.87 19.52 0.13 −0.69
R2 0.9947 0.9921 0.9952
St. Dev 6.45 6.52 0.97 5.34
Mean −0.28 −0.14 −0.05 −0.06
Max 12.90 10.82 2.11 8.98
Min −12.13 −14.81 −3.17 −11.96

VH, fracture toughness and friability can be readily quantified.

Fig. 7. Differences in kWh/t between WVHγ and Wc with respect to P1 (µm).
modelled accumulation effect of grinding resistant materials might be
applicable. This being the case, VH tests could represent a novel method
of determining at a microscopic level an estimation of the specific en-
ergy-particle size relation of ores and blends of materials.
Although it has been shown that the VH of mono-mineralogical,
liberated particles is a major factor determining their BWi, there are
significant differences between the Gγ values and those based on the
result of the division of the test P1 by the root of the VH (GVHγ). It is
suggested that the fracture toughness of materials may be a better in-
dication of their resistance to fragmentation. However, the determina-
tion of the fracture toughness of the small size particles presents tech-
nical difficulties when tested for their BWi. These can probably only be
determined theoretically as reported by Tromans and Meech (2002 and
2004) or by indentation techniques. It is also suggested that although
factors other than VH or fracture toughness such as friability
(Ozkahraman, 2005), internal stresses and planes of weakness, inclu-
sions (type and abundance) affect BWi values for a given material, only
Marshall et al. (2015) have reported that the micro-indentation
method is a rapid, convenient and economical means for comparative
toughness evaluations but that there exists uncertainty in associated
stress intensity factor relations. The similarity of values reported in
distinct investigations such as those of Gong et al. (2001) and Anstis
et al. (1981) for soda-lime glass (0.75. 0.74 and 0.75 MPa m0.5).
Whitney et al. (2007) and Broz et al. (2006) for quartz (1.8 and
1.6 MPa m0.5) suggest that the application of this procedure may be
suitable for the characterization of the energy–size reduction of a given
material. However, compared to the standard VH test, the micro-in-
dentation method is a much slower procedure and it is uncertain
whether or not the time required to determine sufficient measurements
could justify its use.
Line 548 Other established tests for measuring the strength and
elastic properties of ores such as unconfined compression testing, single
impact load tests, etc. have been developed and attempts made to relate
values obtained to their specific grinding energy. To date the relation
between unconfined compression testing and ores BWi has had mixed
results with studies such as that of Doll et al. (2000) not finding any
relation between these values. Studies with single impact load tests
such as described by Tavares and King (1998) having shown a better
relation to specific grinding energies for materials. Free et al. (2005)
have applied this method to develop the Utah Load Cell Work Index
(ULCWI). No comparison of such test results to BWi values adjusted for
sample apparent density and < P1 content appear to have been con-
ducted. The results of such tests to determine the elastic properties of a
material and its relation to specific grinding energy could be processed
as those in this investigation to adjust for sample apparent density
and < P1 content. It would be of significance to determine if such ad-
justments give similar correlations to those reported here. However,
compared to the VH test procedure their application may prove to too
lengthy a process if statistically representative values are required.

234
M. Menéndez et al.
6. Conclusions
Results of BWi tests for five mono-mineralogical samples of diverse
crystallographic nature and VH for P1 values of 500, 250, 125, 980 and
63 µm are presented. Standardizing of the test results for < P1 fines
content and apparent density has been conducted to facilitate analysis
of the relation between specific grinding energy and the grams/re-
volution produced at P1 and VH values.
A lineal equation is obtained relating specific grinding energy and
required corrected Bond Mill revolutions (Revc). This relation indicates
that for a typical BWi test requiring between 29 and 419 mill revolu-
tions, the average grinding energy of the Bond Mill is 194 × 10−7 kWh/
mill revolution. This coincides very closely to the modal average value
reported by Levin (1989) of 198.4 × 10−7 kWh/mill revolution. In
addition, an exponential relation with a perfect correlation factor (R2)
of 1.0 between the increase in specific grinding energy required with
reduction in P1 grinding size is also presented. It is suggested that this
relation is an expression of the increase in the inefficiency of grinding
with reduction in the grinding product size.
It is also shown that the VH of a sample material is a major factor in
controlling the grams/revolution produced at a given P1 size. An ex-
pression with a R2 correlation factor 0.9905 relating the reciprocal of
the root of a sample VH to grams/revolution produced at a given P1 is
presented. This can be applied to estimate a preliminary measure of
final number of mill revolutions required in a BWi test. This value,
when applied to the equation relating the experimental specific
grinding energy to mill revolutions at a given P1 size yields an esti-
mation of the specific grinding energy required for such a grinding
operation. Comparison of the calculated values from Eq. (13) to the
experimental values was found to have an average error of −1.27%.
When the correction for sample apparent densities is applied, the errors
between the measured G and W with those calculated from this equa-
tion must be significantly lower than those indicated here for GVHγ and
WVHγ.
References
Ahmadi, R., Shahsavari, S., 2009. Procedure for determination of ball Bond work index in
the commercial operations. Miner. Eng. 22, 104–106. http://dx.doi.org/10.1016/j.
mineng.2008.04.008.
Aksani, B., Sönmez, B., 2000. Simulation of bond grindability test by using cumulative
based kinetic model. Miner. Eng. 13, 673–677. http://dx.doi.org/10.1016/S0892-
6875(00)00050-9.
Anstis, G.R., Chantikul, P., Lawn, B.R., Marshall, D.B., 1981. A critical evaluation of in-
dentation techniques for measuring fracture toughness: I, direct crack measurements.
J. Am. Ceram. Soc. 64, 533–538. http://dx.doi.org/10.1111/j.1151-2916.1981.
tb10320.x.
Berry, T.F., Bruce, R.W., 1966. A simple method of determining the grindability of ores.
Can. Min. J. 6, 63–65.
Bond, F.C., 1952. The third theory of comminution. Trans. AIME Min. Eng. 193, 484–494.
http://dx.doi.org/10.1016/j.mineng.2006.01.007.
Bond, F.C., 1961. Crushing and grinding calculations, Part II. Br. Chem. Eng. 6, 9–14.
Broz, M.E., Cook, R.F., Whitney, D.L., 2006. Microhardness, toughness, and modulus of
Mohs scale minerals. Am. Mineral. 91, 135–142. http://dx.doi.org/10.2138/am.
2006.1844.
Charles, R.J., 1957. Energy-size reduction relationships in comminution. Min. Eng. 208,
80–88.
Csoke, B., Rácz, Á., Mucsi, G., 2013. Determination of the Bond work index of binary
mixtures by different methods. Int. J. Miner. Process. 123, 78–86. http://dx.doi.org/
10.1016/j.minpro.2013.05.004.
Deniz, V., Ozdag, H., 2003. A new approach to bond grindability and work index: dy-
namic elastic parameters. Miner. Eng. 16, 211–217. http://dx.doi.org/10.1016/
S0892-6875(02)00318-7.
Deister, R.J., 1987. How to determine bond work index using lab ball mill grindability
test. Eng. Min. J. 188, 42–45.
Doll, A.G., Barratt, D., Wood, K., 2000. Comparison of UCS to Bond Work Indices. file:///
C:/Users/Minas/Downloads/UCS_Wi_paper.pdf, (accessed 13.12.2017).
Energy Efficiency Exchange, 2013. Crushing energy costs in the mining sector. accessed
31.3.2016. http://eex.gov.au/2013/11/crushing-energy-costs-in-the-mining-sector/.
235
Minerals Engineering 119 (2018) 228–235
Free, K.S., McCarter, M.K., King, R.P., 2005. Evaluation of a new method for work index
estimation using single particle impact tests. Miner. Metal. Process. 22 (2), 96–100.
Gent, M., Menéndez, M., Torno, S., Toraño, J., Schenk, A., 2012a. Experimental evalua-
tion of the physical properties required of abrasives for optimizing waterjet cutting of
ductile materials. Wear 284–285, 43–51. http://dx.doi.org/10.1016/j.wear.2012.02.
012.
Gent, M., Menendez, M., Toraño, J., Torno, S., 2012b. A correlation between vickers
hardness indentation values and the bond work index for the grinding of brittle
minerals. Powder Technol. 224, 217–222. http://dx.doi.org/10.1016/j.powtec.2012.
02.056.
Gharehgheshlagh, H.H., 2016. Kinetic grinding test approach to estimate the ball mill
work index. Physicochem. Probl. Miner. Process. 52, 342–352.
Gong, J., Chen, Y., Li, C., 2001. Statistical analysis of fracture toughness of soda-lime glass
determined by indentation. J. Non. Cryst. Solids 279, 219–223. http://dx.doi.org/10.
1016/S0022-3093(00)00418-X.
Hukki, R.T., 1962. Proposal for a solomnic settlement between the theories of von
Rittinger. Kick and Bond. Trans. AIME 223, 403–408.
Hukki, R.T., 1975. The principles of comminution: an analytical summary. Eng. Min. J.
176, 106–110.
Kick, F., 1885. Des gesetz der proportionalem wider-stand und seine anwendung. Felizx,
Leipzig.
Levin, J., 1989. Observation on the Bond standard grindability test and a proposal for a
standard grindability test for fine materials. J. South. African Inst. Min. Metall. 89,
13–21.
Magdalinović, N., Magdalinović, S., Jovanovic, R., Stanujkic, D., 2011. Is there a material
constant which characterizes mineral resources grindability? Ann. Univ. Min. Geol.
“St. Ivan Rilski” 54 (I), 116–117.
Marshall, D.B., Cook, R.F., Padture, N.P., Oyen, M.L., Pajares, A., Bradby, J.E., Reimanis,
I.E., Tandon, R., Page, T.F., Pharr, G.M., Lawn, B.R., 2015. The compelling case for
indentation as a functional exploratory and characterization tool. J. Am. Ceram. Soc.
98, 2671–2680. http://dx.doi.org/10.1111/jace.13729.
Morrell, S., 2004. An alternative energy-size relationship to that proposed by bond for the
design and optimisation of grinding circuits. Int. J. Miner. Process. 74, 133–141.
http://dx.doi.org/10.1016/j.minpro.2003.10.002.
Mosher, J.B., Tague, C.B., 2001. Conduct and precision of bond grindability testing.
Miner. Eng. 14, 1187–1197.
Namura, S., Tanaka, T., 2011. Analysis of energy–size reduction relationships in batch
tumbling ball mills. Powder Technol. 208 (3), 610–616. http://dx.doi.org/10.1016/j.
powtec.2010.12.028.
Ozkahraman, H.T., 2005. A meaningful expression between bond work index, grind-
ability index and friability value. Miner. Eng. 18, 1057–1059. http://dx.doi.org/10.
1016/j.mineng.2004.12.016.
Ozkahraman, H.T., 2010. Breakage mechanisms and an encouraging correlation between
the bond parameters and the friability value. J. South. African Inst. Min. Metall. 110,
153–159.
Tromans, D., Meech, J.A., 2004. Fracture toughness and surface energies of covalent
minerals: theoretical estimates. Miner. Eng. 17, 1–15. http://dx.doi.org/10.1016/j.
mineng.2003.09.006.
Tavares, L.M., Kallemback, R.D.C., 2013. Grindability of binary ore blends in ball mills.
Miner. Eng. 41, 115–120. http://dx.doi.org/10.1016/j.mineng.2012.11.001.
Tavares, L.M., King, R.P., 1998. Single-particle fracture under impact loading. Int. J.
Miner. Process. 54 (1), 1–28. http://dx.doi.org/10.1016/S0301-7516(98)00005-2.
Tromans, D., Meech, J.A., 2002. Fracture toughness and surface energies of minerals:
theoretical estimates for oxides, sulphides, silicates and halides. Miner. Eng. 15,
1027–1041. http://dx.doi.org/10.1016/S0892-6875(02)00213-3.
von Rittinger, P.R., 1987. Lehrbuch der aufbereitungs kunde. Ernst and Korn, Berlín.
Whitney, D.L., Broz, M., Cook, R.F., 2007. Hardness, toughness, and modulus of some
common metamorphic minerals. Am. Mineral. 92 (2–3), 281–288. http://dx.doi.org/
10.2138/am.2007.2212.
Zambrano, A.R., 2002. General work input for sizing grinding ball mills. Miner. Metall.
Process. 19, 178–186.
Further reading
ASTM, 2012. ASTM E384: Standard Test Method for Knoop and Vickers Hardness of
Materials ASTM Standards. pp. 1–43. 10.1520/E0384-11E01.2.
ASTM, 2003. ASTM E276-03, 2003. Standard Test Method for Particle Size or Screen
Analysis at No, 4 (4,75-mm) Sieve and Finer for Metal-Bearing Ores and Related
Materials. ASTM Standards. 10.1520/E0276-03.
Magdalinović, N., 1989. A procedure for rapid determination of the bond work index. Int.
J. Miner. Process. 27, 125–132. http://dx.doi.org/10.1016/0301-7516(89)90010-0.
Magdalinović, N., 2003. Abbreviated test for quick determination of Bond’s work index. J.
Min. Metall. A Min. 39, 1–10.
Mucsi, G., 2008. Fast test method for the determination of the grindability of fine ma-
terials. Chem. Eng. Res. Des. 86, 395–400. http://dx.doi.org/10.1016/j.cherd.2007.
10.015.