Content uploaded by F. Wall
Author content
All content in this area was uploaded by F. Wall on Apr 09, 2018
Content may be subject to copyright.
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/authorsrights
Author's personal copy
The influence of equipment settings and rock properties on high voltage breakage
Klaas Peter van der Wielen
a,
⇑
, Richard Pascoe
a,
⇑
, Alex Weh
b
, Frances Wall
a
, Gavyn Rollinson
a
a
Camborne School of Mines, University of Exeter Cornwall Campus, Penryn, Cornwall TR10 9EZ, United Kingdom
b
SelFrag AG, Kerzers CH-3210, Switzerland
article info
Article history:
Received 30 May 2012
Accepted 11 February 2013
Keywords:
Comminution
High voltage pulse breakage
SELFRAG
Electric pulse disaggregation
Electrical comminution
Energy efficiency
abstract
High voltage breakage is a novel comminution method that relies on highly energetic electrical pulses to
weaken or fully fragment rocks. The potential of this technology to improve liberation and increase the
grindability of ores has been demonstrated previously, but the fragmentation process is not fully under-
stood. In this study a total of 20 rock types were treated in a SELFRAG Lab device to determine the influ-
ence of equipment parameters on breakage. Rock mass properties and Bond Work Index were determined
for each rock type to identify their relation to breakage behaviour. Results show how, by influencing total
applied energy, the number of discharges and voltage are the two major influences on the resultant prod-
uct size. It has also been shown that coarser feed sizes are more amenable to high voltage breakage.
Acoustic impedance, porosity and quartz content were found to relate to breakage but Bond Work Index
only correlates loosely.
Ó2013 Elsevier Ltd. All rights reserved.
1. Introduction
Tromans (2008) and Norgate and Jahanshahi (2011) have high-
lighted the energy inefficiency of comminution processes. Improv-
ing comminution efficiency has been the focus of much research in
recent years and initiatives with this goal include a platform to dis-
cuss energy use in comminution and raising awareness of the issue
(Coalition for Eco-Efficient Comminution), extensive scientific re-
search aiming to optimise energy utilisation of existing processes
and development of new technologies.
High voltage discharges to pre-weaken or fully fragment rocks
offers a new technology with considerable potential. It is a rela-
tively novel comminution technique that may improve energy util-
isation in comminution through improved liberation, less fines
generation and weakening of rocks prior to grinding (Wang et al.,
2011, 2012). The technology relies on inducing electrical break-
down, which occurs when the resistivity of a dielectric is insuffi-
cient to completely block all transfer of electricity, whilst
conductivity is not high enough to fully accommodate this flow
of electricity without considerable changes to the crystal lattice.
The Marx generator, crucial for development of high voltage
pulses, was invented in 1924 by Erwin Otto Marx. However, it
was not until the Cold War era that Russian scientists realised its
potential in mineral processing, after a chance discovery that high
voltage discharges in water generated shockwaves powerful en-
ough to crush rock (Andres, 2010). This form of high voltage break-
age (characterised by slower pulse rise-times), better known as
electro-hydraulic crushing, was later superseded by the more effi-
cient electro-dynamic technology under investigation in this pa-
per, which uses faster pulse rise-time electrical discharges to
induce electrical breakdown in the rock rather than in water. For
a more in-depth description of the early history and evolution of
high voltage breakage technology, readers are referred to Andres
(2010).
In the 1990s, several research institutions, including the Fors-
chungszentrum Karlsruhe (FZK), Germany and Imperial College,
London, embarked on research programs investigating high volt-
age breakage technology, its potential applications and commercial
prospects. Mineral processing applications were the focus of the
work at Imperial College, whilst FZK concentrated on industrialisa-
tion of high voltage breakage products in a variety of specific appli-
cations. In 2007 SELFRAG acquired licences for the technology from
FZK and embarked on an extensive research and development pro-
gramme to market high voltage equipment for the minerals and
materials processing industries. Parallel to research at FZK and
Imperial College London, a consultancy (CNT-MC) based in Canada
also carried out research (e.g. Rudashevsky et al., 1995, Lastra et al.,
2003) into the technology.
Interest in this technology has increased significantly in recent
years, whilst most work has focussed on proof-of-concept, with lit-
tle systematic investigations into underlying processes. Andres
et al. (2001a,b) and Wang et al. (2011, 2012) have demonstrated
the potential of high voltage breakage as a mineral processing
technology. Andres et al. (2001a,b) and Wang et al. (2012) focused
on using high voltage breakage technology for full fragmentation,
whereas Wang et al. (2011) focused on using the technology to
0892-6875/$ - see front matter Ó2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.mineng.2013.02.008
⇑
Corresponding authors. Tel.: +44 1326 371838.
E-mail addresses: kpv203@ex.ac.uk (K.P. van der Wielen), r.d.pascoe@exeter.
ac.uk (R. Pascoe).
Minerals Engineering 46-47 (2013) 100–111
Contents lists available at SciVerse ScienceDirect
Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng
Author's personal copy
weaken rocks. Andres et al. (2001a) reported improvements in
grades and/or recoveries after high voltage treatment for copper,
nickel and platinum-group element ores. Specific energy inputs
for the different tests were not reported, but in a subsequent paper
Andres et al. (2001b) reported energy consumption was nearly two
times as high as that for mechanical comminution of the same ore
(90 vs. 50 kW h t
1
). In a more recent publication, Andres (2010)
provided a compilation of other promising results on flotation
and leaching behaviour of fragmented ores to highlight the poten-
tial of high voltage breakage technology. Wang et al. (2012) per-
formed an extensive investigation using a Mineral Liberation
Analyser (MLA) to compare liberation from high voltage breakage
and mechanical comminution at the same specific energy input.
They found high voltage breakage produced significantly coarser
products with better liberation of the minerals of interest. They
suggested there may be an optimal range of specific energy appli-
cation in high voltage breakage that would yield the required lib-
eration of target minerals. A different approach was taken by
Wang et al. (2011), who used a Rotary Breakage Tester (JKRBT) to
determine product residual hardness after high voltage treatment
of between 1 and 3 kW h t
1
. They found a weakening of 9–52%
after treatment and predicted energy savings of up to 24% during
the comminution process. It is unknown whether improved liber-
ation such as that found by Wang et al. (2012) is still available at
these lower energy inputs. The positive energy balance of the
weakening approach has made it the preferred route for using high
voltage breakage in processing circuits. Construction of a 100t/h
demonstration plant is underway and the ultimate goal is manu-
facturing continuous equipment with industrial-scale (>500 t/h)
throughput for weakening of rocks.
Fragmentation processes during high voltage breakage are
distinctly different from those in mills and crushers. According to
Andres et al. (2001b) plasma channels form within the dielectric
solid during electrical breakdown. These channels undergo explo-
sive radial expansion, giving rise to powerful shockwaves that re-
sult in micro-crack formation and ultimately may cause
fragmentation of the solid. Bluhm et al. (2000) also postulated that
the high voltage-induced fragmentation process relies largely on
shock waves and induced tensile forces, causing breakage at
inhomogeneities when tensile strength is exceeded, but experi-
mental data to support this is limited. However, in an earlier paper,
Andres et al. (2001a) attributed micro-crack generation and min-
eral liberation to plasma capillaries, implying that there is a signif-
icant component of direct fragmentation action by the plasma. This
dichotomy demonstrates that further experimental work is needed
to clarify high voltage breakage processes. In addition, little data
have been published on the effect of equipment settings on frag-
mentation, selective fragmentation and weakening, and how these
factors interact with rock properties. These data are essential in
designing effective processing configurations and protocols and
may provide limitations to the use of high voltage breakage equip-
ment in certain applications.
This paper aims to clarify some of these aspects of the technol-
ogy and aid optimisation of process variables. This was done by
investigating the influence of equipment settings and feed size
on the resultant product after high voltage treatment. Other factors
such as selective fragmentation, fines reduction and product shape
are not discussed further. The focus of this paper is on the effects of
voltage and the total number of discharges, as these are the two
ways of varying total energy input, but some observations relating
to electrode gap and pulse rate will also be discussed. During the
tests it was also found that there was a strong effect of feed size
on fragmentation behaviour. With current work to scale up high
voltage breakage technology in mind, results presented in this pa-
per will be discussed in terms of up-scaling and integration of the
technology into a processing circuit.
2. Methods
2.1. Feed material
The rock types used as feed material in this study were selected
to give a broad spectrum in properties and genetic origins. Initial
work was carried out on rock types used for aggregate production
because of their relatively simple and uniform mineralogy. Sulp-
hides in particular tend to have more variable electrical properties
(mainly permittivity) when compared to common rock-forming
minerals and at the outset of the study it was uncertain how this
would interact with breakage behaviour in the high voltage re-
gime. To minimise any potential complications due to the presence
of sulphides, these rocks were used as the feed material for initial
experiments. Later tests on mineral ores were performed to com-
pare and contrast their behaviour to that of the aggregates and
to reinforce the link to mineral processing applications.
Most rocks were obtained as lump material to allow core
extraction for geomechanical testing. The remainder of the rocks
were stage crushed to the desired feed size. The feed size for most
tests was 20 + 14 mm, with the exception of the material for
those tests where feed size was the variable under investigation.
Aggregates were checked for uniformity and any rocks of unusual
appearance were removed by automated optical sorting.
2.2. High voltage treatment
The high voltage treatments were performed in batches in a
SELFRAG Lab unit, manufactured by SELFRAG AG, Switzerland
(Fig. 1). The device relies on a transformer feeding a Marx genera-
tor to generate pulses and discharge them into a process vessel.
This process vessel sits on a lifting table that moves it into a
shielded processing area. The process vessel can be used in open
and closed configuration. The closed configuration utilises a closed
bottom electrode as opposed to the open configuration where the
bottom electrode is basically a sieve deck with a separate sample
collection chamber underneath. The open configuration allows
material of the desired grain size through, which prevents them
from further disintegration and using spark energy for breakage
beyond the target size.
In the SELFRAG Lab unit, the voltage (90–200 kV), electrode gap
(10–40 mm), pulse rate (1–5 Hz) and number of electrical pulses
(1–1000) can be varied, and it is designed for batch processing of
samples of up to approximately 1 kg. For this research the high
voltage treatment was carried out on batches of 700 g in a closed
vessel with de-mineralised water as the medium (conductivity
<10
l
s/cm).
Prior to initial testing, standard test settings were defined in
conjunction with SELFRAG. The number of discharges was set at
300 with a voltage of 140 kV, a pulse rate of 3 Hz and with an elec-
trode gap of 25 mm. At these settings, all particles are affected and
discharges can reliably be achieved. Table 1 outlines the settings
for the different test series. For all tests where the effect of a single
variable was being investigated, three out of four settings were
kept constant whilst the variable of interest was being varied. In
addition, a factorial design experiment was carried out on a por-
phyry copper ore to identify possible interactions between voltage
and number of pulses. In this test, feed size, electrode gap and
pulse rate were kept the same as for the other tests, and 20, 50,
100 and 300 discharges were applied each at 110, 140 and 180 kV.
There is an inefficiency involved in the generation and transfer
of the high voltage pulse which means that the energy consump-
tion by the machine (generator energy) is higher than the energy
discharged into the process vessel (spark energy). This inefficiency
is largely due to several safety features in the lab-scale equipment
K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111 101
Author's personal copy
and the fact the whole setup is geared towards ease of use and
minimal sample loss, rather than process efficiency. As the Lab unit
is a batch processing device, the relatively small throughput means
many of the losses are large compared to what they would amount
to in continuous equipment in terms of energy loss per tonne of
capacity. In addition, many causes of inefficiency can largely be
eliminated through different electrode and Marx generator setups.
Leaving equipment-related factors out of the equation, the
difference between generator and spark energy is dependent on
process water conductivity (high process water conductivity re-
duces discharge ratio), the rock being treated and equipment
settings.
In this paper, reported energy levels refer to the spark energy
input. When assessing the economics or efficiency of high voltage
breakage technology it is recommended that generator energy in-
put is relied upon, but for this research spark energy input is pre-
ferred as it leaves out of the equation machine-specific influences
and pulse/discharge inefficiencies.
2.3. Analysis of treated rocks
The main analysis of the high voltage treated products involved
dry sieving on a standard p2 series of sieves from 355
l
mto
45000
l
m. Care was taken to recover the full sample, including
fines, after high voltage treatment. Cubic splines were used to cal-
culate the 80% passing size (P
80
), >14000
l
m fraction and fines
fraction from the available data. Statistical processing of the data,
such as curve regression and dataset comparison, was carried out
using SPSS software. For the latter, the Wilcoxon Signed Ranks test
was used because the assumption of normality required for stan-
dard paired-sample t-tests could not always be guaranteed.
Analyses in this paper focus on product size. Other attributes,
such as selective fragmentation, particle shape after treatment,
fines generation and change in physical properties of the particle
(i.e. weakening) and how these attributes interact with equipment
settings and material properties may also be of interest and are
recommended for consideration in further research.
2.4. Rock properties
Cores were extracted from lump material to facilitate geome-
chanical testing, with a minimum of three cores tested per rock
type. Dimensions and weight were determined accurately for each
core to give density (kg m
3
), and sound velocity (m s
1
) of the
rocks was measured using a Posso acoustic tester. These two values
were used to calculate acoustic impedance (i.e. Z=
q
C
0
, where
q
is
the density and C
0
is the speed of sound in a material). These cores
were then used to determine average uni-axial compressive and
tensile strengths (Brazilian test) in a rigid load frame. Young’s
Modulus was determined from the load profile obtained during
the compressive strength tests. Irregularly shaped particles and
off-cuts from cores were used for determination of the Point Load
Index. For all geomechanical test work, guidelines by the Interna-
tional Society for Rock Mechanics were followed (ISRM, 1981).
The Bond Ball Mill Work Index was determined following a guide-
line by Deister (1987) at a closing size of 355
l
m, with the sample
for these tests derived from the same sample batch or lump mate-
rial as the rock treated in the SELFRAG Lab unit.
Quartz and sulphide content, porosity and characteristic grain
size were determined from QEMSCAN analysis of polished thin sec-
tions. The QEMSCAN 4300 was operated in fieldscan mode, running
at an X-ray pixel spacing of 10
l
m in conjunction with iMeasure
v4.2 and iDiscover v4.2 and v4.3 software for data acquisition
and processing. General operational procedures for sample prepa-
ration and data processing/analysis, as outlined in Pirrie et al.
(2004) and Rollinson et al. (2011), were followed. Quartz and sul-
phide content are determined during X-ray analysis and the poros-
ity was calculated by classing internal background, glass and resin
within a sample as porosity (injector function). Data from QEM-
SCAN investigation are considered a semi-quantitative indication
because of the 2-dimensional nature of the sample measured, the
relatively limited amount of data and potential stereological errors.
Furthermore, given the 10
l
m X-ray pixel spacing during analysis,
it is only relevant to the >10
l
m portion of porosity. Armitage et al.
(2010) report a comparison of porosity data from QEMSCAN and
mercury porosimetry, showing that QEMSCAN can be used to
Transformer
Marx Generator
Process vessel with
sample submerged
in water
Discharge electrode
Ground electrode
Lifting table
Fig. 1. Schematic of SELFRAG Lab unit.
Table 1
Equipment settings for the different test setups.
Test variable Voltage (kV) No. of discharges Electrode gap (mm) Pulse rate (Hz)
Voltage 90–200 300 25 3
No. of discharges 140 5–850 25 3
Electrode gap 140 300 10–40 3
Pulse rate 140 300 25 1–5
Factorial design 110, 180 20, 50, 100, 300 25 3
Feed size 140 300 25 3
102 K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111
Author's personal copy
determine porosity, though it does tend to underestimate it. The
QEMSCAN iDiscover software calculates average grain size by add-
ing up the total length of all the horizontal intercepts for a mineral
grain measured in a sample and divides by the number of inter-
cepts for that grain to give an average grain size per mineral. From
the QEMSCAN data, characteristic grain size for the whole sample
was calculated by weighting the reported grain size of a mineral
by its mineral volume as reported by iDiscover software.
3. Results and discussion
3.1. Equipment settings
3.1.1. Number of discharges and voltage
The voltage applied to a sample determines the amount of en-
ergy deposited per discharge and each discharge represents an
incremental amount of additional spark energy. For purposes of
clarity it is important to make a distinction between pulses (every
electrical burst of energy generated by the Marx generator) and
discharges (only those pulses that induce electrical breakdown in
the rock sample). The SELFRAG Lab unit can be set to produce
any discrete number of electrical pulses, but not each electrical
pulse develops into a discharge affecting the rock (i.e. not all pulses
induce electrical breakdown in the rocks).
Fig. 2 shows the influence of the total number of applied dis-
charges on the product size after high voltage treatment. Invari-
ably, each rock type exhibited an initial phase (up to
approximately 75 discharges or 7 kW h t
1
) where little size reduc-
tion occurred, followed by a strong decrease in product size over a
relatively small energy range, before levelling off in the high en-
ergy range. This trend was observed for every single rock type.
The influence of voltage on product size is illustrated in Fig. 3.An
important observation is that, especially at lower voltages
(<120 kV), a large number of pulses may be needed to achieve
the desired number of discharges (up to 1800 pulses to achieve
300 discharges). For some rock types discharges can easily be
achieved at voltages as low as 90 kV (hornfels, quartz monzodior-
No. of discharges
10008006004002000
Whole sample P80
(µm)
20,000
15,000
10,000
5,000
0Tuff
Soapstone
Slate
Shale/Massive Sulphide (Au ore)
Sandstone
Quartz Monzodiorite (Au ore)
Pegmatite (Ta/Li ore)
Metagabbro
Limestone
Iron ore (BIF)
Hornfels
Granodiorite (Au ore)
Granite (porphyritic)
Granite (medium-grained)
Granite (fine-grained)
Gneiss
Dolerite
Chert
Andesite (porphyry Cu ore)
Altered Metagabbro
Fig. 2. Whole sample product size as a function of the number of discharges. Voltage (140 kV), electrode gap (25 mm), feed size (20 + 14 mm) and pulse rate (3 Hz) kept
constant.
Voltage (kV)
20018016014012010080
Whole sample P80
(µm)
20,000
15,000
10,000
5,000
0
Tuff
Shale/Massive Sulphide (Au ore)
Quartz Monzodiorite (Au ore)
Metagabbro
Limestone
Hornfels
Granite
Dolerite
Andesite (porphyry Cu ore)
Fig. 3. Whole sample product size as a function of the voltage. No. of discharges (300), electrode gap (25 mm), feed size (20 + 14 mm) and pulse rate (3 Hz) kept constant.
K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111 103
Author's personal copy
ite) whereas others need voltages in excess of 110 kV to achieve
any discharges (granite, tuff, dolerite).
The number of discharges and the applied voltage are both di-
rectly proportional to the energy input into a sample during high
voltage breakage. An important question therefore is whether
treatment at different voltages but similar total spark energy in-
puts yields comparable particle size distributions. Fig. 4 combines
data from Figs. 2 and 3 and shows that both distributions display
similar trends, though the voltage tests span a smaller energy
range (as a consequence of test setup). Further statistical analysis,
comparing the measured and predicted product size showed that
in a case-by-case analysis on six rock types, only one (quartz mon-
zodiorite) showed significant deviation from the predicted product
size.
To further examine the effect of voltage, factorial design-style
experiments were done on a porphyry copper ore and a gold-
bearing quartz monzodiorite, with different combinations of dis-
charges (20, 50, 75, 300) at three voltages (110, 140 and 180 kV).
Figs. 5 and 6 shows the result from these tests. For both ores there
is no significant difference between particle size distributions after
treatment at 140 and 180 kV. However, for both test series the
highest energy 110 kV test had a significantly coarser particle size
distribution than predicted for 140 kV and 180 kV treatments.
Moreover, when comparing tests for these two rock types where
the voltage was the sole variable of interest, the particle size distri-
butions from treatments below 130 kV all yielded coarser products
than expected. This behaviour is most pronounced in the monzodi-
orite but can also be observed to a lesser extent in the porphyry
copper ore.
The discharges–voltage comparison results suggest that total
applied energy is the main variable to consider for product size dis-
tribution, but that at lower voltages the rate at which particles get
broken out of the feed fraction may be lower. The applied voltage
governs energy per discharge, and it may be that a ‘threshold’ volt-
age is required to fully overcome particle strength and directly
cause breakage. Below this threshold voltage particles still accrue
incremental damage but it may take multiple discharges to disin-
tegrate particles enough to make them report to a size fraction be-
low that of the feed. The monzodiorite has a comparatively high
tensile strength, which may contribute to this behaviour but more
detailed investigations are recommended to ascertain the cause of
the effect. It is also inconclusive whether certain voltage/discharge
combinations yield more or less pronounced pre-weakening and if
liberation is affected.
3.1.2. Total energy input
No general relationship could be defined that described product
sizes from both breakage over the entire energy input range, so the
data were separated into two distinct datasets. The first considered
the percentage of feed size remaining (i.e. the weight of the
>14 mm fraction); the second dataset considered the particle size
distribution of the product.
Spark energy input (kWh t-1)
806040200
Whole sample P80
(µm)
20,000
15,000
10,000
5,000
0
Voltage
No. of discharges
Fig. 4. Comparison of whole sample product size after treatment with voltage or
no. of discharges as main variable under investigation. Electrode gap (25 mm) and
pulse rate (3 Hz) kept constant.
Spark energy input (kWh t-1)
50403020100
Whole sample P80
(µm)
20,000
15,000
10,000
5,000
0
180 kV
140 kV
110 kV
Andesite (porphyry Cu ore)
Fig. 5. Whole sample product size as a function of spark energy input applied
through different voltages for a porphyry copper ore. Electrode gap (25 mm) and
pulse rate (3 Hz) kept constant.
Spark energy input (kWh t-1)
50403020100
Whole sample P80
(µm)
20,000
15,000
10,000
5,000
0
180 kV
140 kV
110 kV
Quartz Monzodiorite (Au ore)
Fig. 6. Whole sample product size as a function of spark energy input applied
through different voltages for a gold ore. Electrode gap (25 mm) and pulse rate
(3 Hz) kept constant.
104 K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111
Author's personal copy
Fig. 7 shows the decrease in weight percentage of >14 mm par-
ticles with spark energy input (W). The rate of decrease followed
an exponential relationship for every rock type tested:
%of feed remaining ¼C
e
exp
Se W
ð1Þ
where S
e
is a material-dependent exponent that determines the
steepness of the function describing breakage of the >14 mm frac-
tion and C
e
is a constant. For most rock types the constant was close
to 1 (i.e. no particles in the <14 mm fraction at no energy applied),
so the constant is not included in this relationship. The majority of
particles in the coarsest size fraction (>14 mm) showed little or no
sign of being affected by high voltage pulses. As energy inputs in-
crease (i.e. more discharges applied), the probability of a particle
being affected by a discharge increases, and this may be reflected
in the exponential nature of the distribution. However, the slope
of this relationship was also found to vary significantly between
rock types and Section 3.3 outlines efforts to relate feed size break-
age to rock properties. With the exception of two rock types (insuf-
ficient data in the <20 kW h t
1
range), the correlation coefficient
for each rock type exceeded 0.80 (mostly >0.95) and all fits were
highly significant (p< 0.01).
The 80% passing size of the product (<14 mm) for all 20 investi-
gated rock types are shown in Fig. 8. At energy levels above
5kWht
1
the decrease of particle size distribution with spark en-
ergy input can be described by a power/fractal law:
Product P
80
¼C
f
W
Sf
ð2Þ
where C
f
is the rock-specific constant, and S
f
is the rock-specific
exponent.
Table 2 lists the fitted parameters Se and Sf for each of the
tested rock types. Each rock-specific relationship had a correlation
coefficient above 0.85 and the significance was below 0.01. Links
between these rock-specific energy–size relationships and rock
properties are discussed in Section 3.3. Between 0 and circa
5kWht1 of spark energy applied, there is a phase where size
reduction does not fit the power/fractal law. This is thought to re-
flect an initial breakage phase where the majority of particles have
not yet accrued sufficient damage to report to the <14 mm fraction.
When affected by a discharge, micro-cracks are formed in a particle
and some spalling may occur. Once enough energy has been ap-
plied (and depending on particle properties), micro-crack density
is thought to increase sufficiently to produce an interconnected
fracture network, eventually reducing a particle’s integrity to the
point where it fragments completely.
The amount of generator energy released by the Marx generator
can be calculated accurately from the applied voltage and the num-
ber of discharges, regardless of rock type. The amount of spark en-
ergy transferred to in a sample can also be accurately calculated
from these two variables, but there is considerable variation be-
tween rock types, which suggests the conversion of generator to
spark energy is rock-specific. Generator-spark energy conversion
ratios were observed ranging from 80% to 90% range (sandstone,
iron ore) to <60% (soapstone) and were used as a measure of the
efficiency of the conversion process.
3.1.3. Electrode gap and pulse rate
The other two equipment settings that can be varied in the SEL-
FRAG Lab unit are the electrode gap (distance between discharge
and ground electrode in the processing vessel) and pulse rate
(no. of pulses per second).
It was found that at certain processing parameters (typically
low voltage gradient/high processing water conductivity), a por-
tion of pulses from the Marx generator did not develop into dis-
charges. Generation of each pulse consumes a fixed amount of
energy, regardless of whether a discharge is developed and conse-
quently every ‘misfired’ pulse (i.e. no discharge) represents lost
energy.
By definition, the discharge ratio (no. of discharges divided by
number of pulses) cannot exceed 1 and in the available dataset it
ranged from 1 down to 0.6 for most rock types tested. The conver-
sion of generator energy to spark energy (i.e. spark energy divided
by generator energy) was used as an indicator of electrical effi-
ciency of high voltage breakage. Fig. 9 shows at discharge ratios
smaller than 0.95, a linear decrease (r
2
= 0.84, sig. < 0.001) of elec-
trical efficiency was observed with a decrease in the discharge ra-
tio. Therefore, this ratio is a key factor to consider in optimisation
of the electrical efficiency of high voltage breakage. Above a dis-
charge ratio of 0.95 the electrical efficiency varied from approxi-
mately 0.6 to >0.9 depending on the rock type.
The electrode gap can influence breakage through two different
routes. Firstly, it governs the volume in the processing vessel avail-
able for particles. This volume accommodates not only the physical
size of the particles, but also their movement. A low electrode gap
may restrict particle movement, resulting in a relatively small
number of particles receiving the bulk of the energy whilst other
Spark energy input (kWh t-1)
3020100
% of >14,000µm particles
100
80
60
40
20
0
Tuff
Soapstone
Slate
Shale/Massive Sulphide (Au ore)
Sandstone
Quartz Monzodiorite (Au ore)
Pegmatite (Ta/Li ore)
Metagabbro
Limestone
Iron ore (BIF)
Hornfels
Granodiorite (Au ore)
Granite (porphyrytic)
Granite (medium-grained)
Granite (fine-grained)
Gneiss
Dolerite
Chert
Andesite (porphyry Cu ore)
Altered Metagabbro
Fig. 7. Mass percentage of particles left in the feed (>14 mm) fraction as a function of total spark energy input for 20 different rock types. Voltage (140 kV), electrode gap
(25 mm) and pulse rate (3 Hz) kept constant.
K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111 105
Author's personal copy
particles are left largely unaffected. Observations suggest the man-
ifestation of the effect of electrode gap on product size is complex.
The product size for some rock types is completely unaffected by
variations in electrode gap, whereas other rock types show varying
degrees of dependence on electrode gap. Where electrode gap was
found to influence product size, the lower electrode gaps (i.e. high-
er voltage gradient but less space for particle movement) yields the
coarser product. This implies the effect is mainly caused by volume
restrictions. Secondly, the electrode gap is of direct influence on
the voltage gradient between electrodes. Electrical breakdown is
a stochastic process and for this to occur the voltage gradient needs
to exceed the electrical breakdown strength of a material (both de-
noted in kV mm
1
). The probability of breakdown occurring in-
creases with voltage gradient till it is close to or at 1. Fig. 10
shows the discharge ratio as a function of voltage gradient. Below
a voltage gradient of 7 kV mm
1
, the discharge ratio was found to
vary strongly between 0.2 and 1. Above this, the discharge ratio
was always larger than 0.95.
The pulse rate was not found to cause significant deviations of
particle size distribution from what was expected from the en-
ergy–size relationship. However, it was observed that a higher
pulse rate made development of a discharge from the high voltage
pulse more probable. Fig. 11 shows this effect for a rock type trea-
ted at 100 kV/25 mm electrode gap, which is near the minimum
voltage gradient required to achieve breakdown for this rock type.
This effect has been observed for other rock types, though its mag-
nitude may vary depending on a rock’s breakdown voltage and the
operating conditions. The SELFRAG Lab unit has a pulse rate range
of 1–5 Hz, so it is unknown whether there is an upper limit to the
influence of pulse rate on the discharge ratio. Plasma effects and
dissipation of electrical charge happen on a much shorter time
scale (<ms), so the pulse rate effect is likely related to residual bub-
bles in the process water after a discharge (Giese and Muller, pers.
comm.). At higher pulse rates these bubbles may not have col-
lapsed fully, or a transient product may still reside in the process-
ing area. The breakdown strength of gaseous phases in these
bubbles is lower than that of water, and therefore they should pro-
vide an alternative path, facilitating transfer of a discharge into the
rock sample that might otherwise not have developed breakdown
in the rock sample.
Spark energy input (kWh t -1
)
806040200
Product P80
(µm)
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0Tuff
Soapstone
Slate
Shale/Massive Sulphide (Au ore)
Sandstone
Quartz Monzodiorite (Au ore)
Pegmatite (Ta/Li ore)
Metagabbro
Limestone
Iron ore (BIF)
Hornfels
Granodiorite (Au ore)
Granite (porphyritic)
Granite (medium-grained)
Granite (fine-grained)
Gneiss
Dolerite
Chert
Andesite (porphyry Cu ore)
Altered Metagabbro
Fig. 8. Product size for the <14 mm fraction as a function of spark energy input for 20 different rock types. Voltage (140 kV), electrode gap (25 mm) and pulse rate (3 Hz) kept
constant.
Table 2
Fitted parameters S
e
and S
f
for all 20 rock types. S
e
for pegmatite not available due to
insufficient data points for a statistically significant fit.
S
e
S
f
Altered metagabbro 0.170 0.957
Andesite 0.131 1.387
Chert 0.158 1.008
Dolerite 0.153 0.933
Gneiss 0.138 0.879
Granite (fine-grained) 0.199 1.477
Granite (medium-grained) 0.259 1.274
Granite (porphyritic) 0.262 1.296
Granodiorite 0.199 1.229
Hornfels 0.175 1.034
Limestone 0.285 1.065
Iron ore (BIF) 0.241 1.217
Metagabbro 0.152 0.891
Pegmatite n/a 1.183
Quartz monzodiorite 0.148 1.213
Sandstone 0.140 3.341
Shale/massive sulphide 0.130 1.402
Slate 0.206 1.634
Soapstone 0.345 1.601
Tuff 0.192 1.146
Discharge ratio
1.000.800.600.400.200
Generator - spark energy conversion efficiency
100%
80%
60%
40%
20%
0%
Fig. 9. Efficiency of conversion of generator to spark energy as a function of
discharge ratio for discharge ratios below 0.95.
106 K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111
Author's personal copy
3.2. Feed size
Fig. 12 shows feed size (represented through mean feed size)-
was shown to have a very distinct effect on reduction ratio
(F
80
/P
80
). This feed size effect is so pronounced that the coarsest
feed sizes invariably produce a finer product size distribution than
the smallest tested feed size after the same total spark energy ap-
plied. This shows that coarser particles are far more susceptible to
high voltage breakage than finer feed sizes, and suggests that the
whole volume of individual particles are affected during high volt-
age breakage. During breakage of coarse particles, the progeny will
at some stage contain particles similar to the finer feed sizes. This
would present a physical limit to size reduction if the feed size ef-
fect occurred mostly concurrent with breakage and result in all
feed sizes producing similar product sizes but this is not the case.
This suggests the particle accrues the damage necessary to produce
the finer size distributions prior to actual size reduction, as other-
wise they would not be comminuted to sizes smaller than that for
smaller feed sizes. Stronger field distortions and more complex
shock wave interactions and reflections in larger particles may be
possible causes for the strong feed size dependence of high voltage
breakage.
Energy transfer may also contribute to the observed feed size ef-
fect. It is conceivable that a larger particle can provide the full bridge
for a discharge from discharge to ground electrode. In this case, all
the energy is deposited in this particle, with a relatively limited tra-
vel distance through the processing water and consequently less
energy loss in the transfer process. Smaller particles on the other
hand will not be able to bridge the gap between electrodes fully
and therefore sparks may be required to jump from particle to par-
ticle several times, involving a longer total travel distance through
water. During this process a larger portion of energy may therefore
be lost in the water and consequently less energy would end up
being available for fragmentation. This mechanism assumes there
is sufficient energy in a 140 kV discharge to cause significant dam-
age in a particle regardless of size as the energy/size ratio would
otherwise favour smaller particles. Validation of the proposed
mechanisms is required as experimental evidence cannot conclu-
sively demonstrate which processes cause the observed effect.
It can also be seen that the feed size effect was evident at vary-
ing magnitudes for each of the five rock types tested. The differ-
ence in product size between the larger feed sizes (larger than
20 + 14 mm) is minimal. The feed size effect has substantial con-
sequences for integration of high voltage breakage technology into
existing processing circuits and will be discussed in more detail la-
ter. In the SELFRAG Lab unit there is a physical limitation to the
largest feed size (approximately 45 mm) that can be fitted into
the SELFRAG Lab processing chamber so it is uncertain what hap-
pens above this size.
3.3. Rock properties
The slope of the exponential relationship describing the weight
percentage of particles left in the feed size (S
e
, see Eq. (1)), and the
slope of the power law describing product size (S
f
, see Eq. (2)) were
used as measures of rock’s response to high voltage breakage. For
curve regression purposes, the positive value of the rock-specific
exponents was used. Tables 3 and 4 list geomechanical and miner-
alogical properties of the rocks used in this research. Fig. 13 shows
the slope of the exponential decrease relationship for feed size as a
function of tensile strength. The general set of data show a linear
relationship between the two variables. The outliers are the sand-
stone (top left) and iron ore (bottom right), and their presence may
be explained by their relatively low and high density. The electrode
gap was not varied in these tests, and therefore the volume of sam-
ple accommodated in the treatment zone between the electrodes
Voltage gradient (kV mm-1)
151050
Discharge ratio
1.00
0.80
0.60
0.40
0.20
0.00
Tuff
Sandstone
Shale/Massive Sulphide (Au ore)
Quartz Monzodiorite (Au ore)
Metagabbro
Limestone
Hornfels
Granite
Dolerite
Andesite (porphyry Cu ore)
Altered Metagabbro
Fig. 10. Discharge ratio as a function of voltage gradient for 11 rock types. No. of discharges (300) and pulse rate (3 Hz) kept constant.
Pulse rate (Hz)
54321
Discharge ratio
1.00
0.80
0.60
0.40
0.20
0.00
Fig. 11. Discharge ratio as a function of pulse rate for quartz monzodiorite. Voltage
(100 kV), no. of discharges (300) and electrode gap (25 mm) kept constant.
K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111 107
Author's personal copy
remained constant. As a consequence, more mass in this ‘hot’ zone
where the majority of discharges travel means more particles will
get affected per discharge. A denser rock means more mass can be
accommodated in the treatment zone and therefore denser rocks
should yield a larger portion of the sample in the product size frac-
tion than expected, with the opposite being the case for low den-
sity rocks. It should be noted that the size and shape of the
treatment zone depends on the electrode geometry. Therefore,
the reported relationship is to some extent specific to the tip-plate
electrode setup in the SELFRAG Lab unit. However, the relationship
to tensile strength is likely to be a generic one irrespective of elec-
trode design. The observed variation between rock types should be
related solely to density (i.e. volume of material in the treatment
zone) if the electrode geometry was the only factor of influence
on particle breakage in the >14 mm fraction, but the correlation
between density and decrease of mass in the feed size fraction is
neither strong nor significant (r
2
0.02, sig. 0.61).
When plotting product size evolution with spark energy (repre-
sented through the slope of the power law) versus rock properties,
it was found that acoustic impedance, Young’s modulus, porosity
and quartz content return good correlations to breakage. Density,
point load index, compressive strength, tensile strength, and char-
acteristic grain size did not show any significant correlation to the
product size evolution with energy input (r
2
< 0.50).
The best fit model for acoustic impedance versus breakage,
shown in Fig. 14, is through a linear model (r
2
0.74,
Table 3
Geomechanical properties of the tested rock types, including standard deviations.
Bond work index
(kW h t
1
)
Compressive
strength (MPa)
Tensile
strength
(MPa)
Young’s
modulus
(GPa)
Density
(kg m
3
)
Point Load
Index (IS
50
)
Acoustic impedance
(Kg s
2
m
1
)
Altered metagabbro 23.4 118 ± 32 13.5 ± 2.6 68.3 ± 4.2 2889 ± 160 11.2 ± 2.2 1.74 10
7
± 1.7 10
6
Andesite
**
17.3 n/a
a
8.1 ± 2.5
b
56.9 ± 17.4
c
2787 ± 42 10.8 ± 3.2 1.24 10
7
± 2.1 10
6
Chert 23.1 116 ± 50 19.6 ± 9.8 59.3 ± 7.4 2695 ± 10 15.1 ± 6.9 1.49 10
7
± 1.3 10
5
Dolerite 24.0 185 ± 46 13.7 ± 4.6 50.9 ± 9.2 2775 ± 36 9.9 ± 2.2 1.66 10
7
± 5.2 10
5
Gneiss n/a 137 ± 38 15.6 ± 3.3 140.1 ± 81.7 2927 ± 30 10.8 ± 3.1 1.84 10
7
± 7.8 10
5
Granite (fine-grained) 11.7 252 ± 21 13.7 ± 1.4 57.4 ± 0.4 2626 ± 3 11.8 ± 2.0 1.43 10
7
± 1.9 10
5
Granite (medium-grained) 14.4 188 ± 10 11.6 ± 1.7 54.5 ± 3.0 2647 ± 10 7.8 ± 1.5 1.44 10
7
± 1.1 10
5
Granite (porphyritic) 14.8 150 ± 3 9.9 ± 1.1 49.1 ± 2.3 2634 ± 5 7.1 ± 1.5 1.37 10
7
± 4.1 10
5
Granodiorite
*
12.8 146 ± 38 11.7 ± 2.9 44.3 ± 10.0 2653 ± 42 11.5 ± 3.0 1.34 10
7
± 8.8 10
5
Hornfels 17.1 227 ± 59 15.4 ± 3.3 58.2 ± 6.7 2871 ± 19 10.7 ± 4.3 1.69 10
7
± 9.0 10
5
Iron ore (BIF) 18.3 136 ± 146 24.0 ± 12.2 63.4 ± 56.2 4021 ± 205 10.0 ± 3.7 2.00 10
7
± 3.0 10
6
Limestone 9.0 152 ± 48 9.1 ± 2.6 64.2 ± 3.8 2710 ± 6 6.0 ± 2.1 1.68 10
7
± 5.6 10
5
Metagabbro 17.5 186 ± 41 15.5 ± 2.1 73.7 ± 1.9 2860 ± 11 10.8 ± 3.4 1.85 10
7
± 3.1 10
5
Pegmatite
***
n/a n/a n/a n/a 2696 ± 21 n/a 1.27 10
7
± 6.6 10
5
Quartz monzodiorite
*
12.3 212 ± 46 15.4 ± 3.0 58.3 ± 6.5 2781 ± 51 10.9 ± 3.0 1.51 10
7
± 1.6 10
6
Sandstone 3.7 107 ± 5
a
6.6 ± 0.4 26.0 ± 1.0 2357 ± 3 4.5 ± 0.6 0.84 10
7
± 7.1 10
4
Shale/massive sulphide
*
13.2 n/a 10.7 ± 4.0
b
46.3 ± 11.8
c
2899 ± 1 8.1 ± 3.1 1.15 10
7
± 1.9 10
6
Slate n/a 167 ± 28 11.0 ± 4.7 50.8 ± 21.0 2789 ± 8 9.7 ± 3.8 1.21 10
7
± 3.6 10
5
Soapstone n/a 88 ± 4 3.4 ± 0.3 25.6 ± 0.4 2839 ± 5 2.3 ± 1.2 0.95 10
7
± 1.6 10
5
Tuff 15.7 105 ± 14 10.0 ± 1.5 47.6 ± 1.4 2706 ± 5 6.8 ± 2.8 1.48 10
7
± 1.4 10
5
*
Gold ore.
**
Copper ore.
***
Tantalum/Lithium ore.
a
Compressive strength estimated from co-linearity with point load index.
b
Tensile strength estimated from co-linearity with point load index.
c
Young’s Modulus estimated from acoustic impedance using Hooke’s Law.
Reduction ratio (F80/P80)
50
40
30
20
10
0
Mean feed size (mm)
50403020100
Quartz Monzodiorite (Au ore)
Limestone
Hornfels
Granite (porphyritic)
Granite (medium-grained)
Granite (fine-grained)
Dolerite
Altered Metagabbro
-45 +31.5 mm
-31.5 +20 mm
-20 +14 mm
-10 +6.3 mm
-14 +10 mm
-6.3 +4 mm
-4 +2 mm
Fig. 12. Reduction ratio as a function of mean feed size for eight different rock types. No. of discharges (300), voltage (140 kV), electrode gap (25 mm) and pulse rate (3 Hz)
kept constant.
108 K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111
Author's personal copy
sig. < 0.001). There is a strong co-linearity between Young’s modu-
lus and acoustic impedance through Hooke’s Law (E=
q
C
02
). The
correlation to Young’s Modulus (r
2
0.70, sig. < 0.001) is not as
strong as that found for acoustic impedance and therefore it is
probably a consequence of the strong influence of acoustic imped-
ance. A low acoustic impedance means a rock is less efficient at
transferring shock wave energy, and hence more energy is ab-
sorbed during the wave transmission process, so the increased size
reduction observed in low acoustic impedance rocks is a logical
finding.
The relation between tensile strength and the disappearance of
particles from the feed size fraction provides experimental evi-
dence for the hypothesis by Bluhm et al. (2000) that fragmentation
occurs in a tensile stress regime. However, the evolution of product
particle size distribution with energy does not correlate well to
tensile strength (r
2
0.22 sig. 0.05). Micro-cracking of rocks is
known to occur during high voltage breakage (Wang et al. 2011).
This may significantly reduce the tensile strength of a rock, and
could explain why the correlation between tensile strength and
product size is not as significant. The correlation between product
size and acoustic impedance provides experimental evidence for
the suggestion that shock wave transmission is a major contributor
to fragmentation during high voltage breakage. This fits in well
with the relation to tensile strength as shock wave reflection and
refraction within inclusions of different acoustic impedance would
give rise to localised tensile stress.
The correlation coefficient between porosity and breakage is
fairly low (0.42) but highly significant (0.005). Porosity is thought
Acoustic impedance (x 107
kg m-2
s-1)
2.52.01.51.00.5
0
Sf
-0.50
-1.00
-1.50
-2.00
-2.50
-3.00
-3.50
Fig. 14. Slope of the product size–energy relationship as a function of acoustic
impedance. Error bars indicate 95% confidence interval.
Table 4
Mineralogical properties of tested rock types, including standard deviations.
Porosity
(%)
Weighted average grain size
(
l
m)
Quartz content
(mass%)
Sulphide content
(mass%)
Carbonate content
(mass%)
Mica content
(mass%)
Altered metagabbro <0.01 134 0.3 <0.1 .1 11.3
Andesite
**
0.13 ± 0.08 37 ± 6 16.1 ± 5.7 4.8 ± 3.0 0.4 ± 0.3 31.9 ± 9.1
Chert <0.01 54 47.6 1.08 <0.1 17.2
Dolerite <0.01 39 13.1 <0.1 7.9 49.9
Gneiss 0.17 101 <0.1 <0.1 <0.1 5.4 ± 1.9
Granite (fine-grained) 0.08 ± 0.02 140 ± 5 32.0 ± 1.2 <0.1 <0.1 9.5 ± 0.3
Granite (medium-
grained)
0.42 ± 0.56 248.4 ± 75 31.4 ± 7.1 0.1 ± 0.03 <0.1 12.5 ± 3.1
Granite (porphyritic) 0.12 ± 0.04 264 ± 56 35.0 ± 5.7 <0.1 <0.1 12.3 ± 0.1
Granodiorite
*
0.47 ± 0.60 94 ± 25 27.3 ± 3.9 1.8 ± 0.5 4.5 ± 0.1 9.2 ± 1.9
Hornfels <0.01 39 0.6 <0.1 <0.1 29.5
Limestone 0.26 n/a <0.1 <0.1 99.6 <0.1
Iron ore (BIF) 7.42 ± 1.06 n/a n/a n/a n/a n/a
Metagabbro <0.01 154 <0.1 <0.1 <0.1 11.1
Pegmatite
***
n/a n/a 80.6 <0.1 0.1 0.9
Quartz monzodiorite
*
<0.01 84 6.0 4.9 27.6 21.2
Sandstone 6.31 103 ± 3 60.4 ± 1.7 <0.1 6.4 ± < 0.1 4.6 ± 0.6
Shale/massive
sulphide
*
0.59 ± 0.22 305 ± 347 37.0 ± 25.4 21.6 ± 17.9 29.5 ± 22.5 8.1 ± 7.2
Slate 0.107 23 15.4 0.3 1.6 46.1
Soapstone 0.55 48 <0.1 0.35 15.5 28.3
Tuff <0.01 41 ± 1 15.6 ± 1.1 <0.1 6.1 ± 1.2 46.8 ± 1.3
*
Gold ore.
⁄⁄
Copper ore.
***
Tantalum/Lithium ore.
Se
-.10
-.20
-.30
-.40
Tensile strength (MPa)
40.0
30.0
20.0
10.0
0.0
Fig. 13. Slope of the exponential decrease of mass in the feed size (>14 mm)
fraction as a function of tensile strength. Error bars indicate 95% confidence interval.
K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111 109
Author's personal copy
to be of importance because the two major outliers in Fig. 14
(sandstone and iron ore), have the highest porosities (5.62% and
6.36% respectively) of all rock types tested (average porosity
0.22% for the rest of the data set). This may be related to the lower
electrical breakdown strength of air, which invariably is lower than
that of water and rocks regardless of pulse rise time and voltage
(Andres et al., 2001b). Air trapped in pores may therefore be more
facilitating to electrical breakdown and the formation of a plasma
channel, resulting in more efficient breakage. Further investigation
of this effect is recommended due to the relatively limited distribu-
tion of porosity (0.001–0.5% when excluding the two outliers) for
the available rock types. Though permeability was not considered
in this research, it may also be of influence on breakage by allowing
treatment water to percolate into voids occupied by prior to high
voltage treatment. This could to some extent negate the positive
effect of porosity on the ease of fragmentation during high voltage
breakage.
The quartz content appears to be related to breakage through a
linear relationship (r
2
0.54, sig. 0.004). Furthermore, the finest
product sizes at any given treatment were observed for the rock
types with the highest quartz content materials (pegmatite and
sandstone). It is possible the influence of quartz content is related
to piezo-electric behaviour (i.e. charge accumulation in response to
mechanical stress) of quartz, or its brittle nature but no conclusive
explanation is yet available. Co-linearity of porosity and quartz
content with other properties such as acoustic impedance is not
strong enough to explain the influence of these properties.
No correlation could be established between measured rock
properties and the minimum voltage gradient required for break-
down. Likewise, the rock-specific variation in generator to spark
energy conversion could not be explained by any known rock prop-
erties. As both are related to electrical characteristics of high volt-
age treatment they are more likely linked to electrical properties
and these values were not available in this research. Further work
is underway to determine interaction between high voltage pro-
cessing and the electrical properties of the rock being processed.
The correlation between Bond Work Index and the evolution of
product size with energy input (Fig. 15) is significant and strong
(r
2
0.89, sig. < 0.001), but heavily reliant on the sandstone outlier
(r
2
0.20, sig. 0.11 when sandstone is excluded). Therefore,
Bond Work Index may serve as a very rough indicator of ease of
breakage during high voltage treatment, which may prove useful
given the fact that a Work Index is determined for practically every
ore being comminuted. At the same time though, it should be
pointed out that the variation in Bond Work Index between sam-
ples is far larger than that observed in ease of breakage by high
voltage discharges. A good example is the hornfels producing
nearly exactly the same product size as the limestone, despite hav-
ing a Bond Work Index almost twice as high (17.1 vs. 9.0 kW h t
1
).
Implications of these observations will be discussed further in the
following section.
4. Relevance of findings to a continuous process
Continuous high voltage breakage equipment for weakening of
ores should have two primary goals: (1) to achieve an optimal bal-
ance between energy introduced into an ore and reduction in en-
ergy requirements due to high-voltage induced weakening of
ores, and (2) to apply the pulsed energy in the most efficient
manner.
The data presented in this paper show fragmentation behaviour
is rock-specific. Figs. 5, 6 and 8 show that high voltage energy in-
puts generally are too high to make it a feasible technology for full
fragmentation unless the improved liberation (and potentially bet-
ter grade/recovery), such as that reported by Wang et al. (2012)
can be used to justify the higher energy input. Individual assess-
ment per rock type will be required to determine where the opti-
mal trade-off is between reduced energy demand after
weakening and high voltage energy spent in achieving this weak-
ening. Research by Wang et al. (2011), and initial research results
available at the Camborne School of Mines suggest that a signifi-
cant reduction in energy requirements after weakening can be
achieved at energy inputs of approximately 2–5 kW h t
1
of spark
energy.
On the basis of presented data it is impossible to pin-point a
particular combination of voltage and total number of pulses to
achieve an optimal trade-off between weakening and high voltage
energy input. However, the data do show conclusively that the
electrode gap at a pre-selected voltage (i.e. the voltage gradient)
should be high enough to exceed the threshold value where the
discharge ratio as high as possible (i.e. >0.95). Results suggest
7kVmm
1
should be sufficient, regardless of other variables such
as process water conductivity. At the same time, electrode gap also
influences throughput by determining the volume available be-
tween electrodes, and hence the top size of particles that can be
treated. This means a compromise has to be considered when
increasing the voltage gradient at the expense of top size treated.
Individual assessment of the most suitable feed size, voltage and
electrode gap for a rock type is likely needed to determine ideal
settings. Because this entire publication is based on data from a
small scale batch process, it is strongly recommended that selected
experiments are reproduced on a larger scale or in a locked-cycle
test to determine unit performance in a continuous processing
environment, especially with regards to energy utilisation.
A higher pulse rate means the same amount of energy can be
applied in a shorter period of time, and hence the residence time
of particles in the treatment area can be reduced. At the same time
it also increases the likelihood of a pulse developing a discharge so
for efficiency purposes it is recommended the pulse rate is main-
tained as high as possible.
With regards to rock mass properties, it appears rock types with
any combination of low acoustic impedance, high porosity and
high quartz content are most amenable to high voltage breakage.
It is recommended these materials are targeted for high voltage
breakage experiments. If rock cores are available (i.e. diamond drill
core on an exploration project), acoustic impedance can be deter-
mined in a time and cost effective manner using precision scales,
a vernier calliper and an acoustic tester. The presence of silica in
a large number of common rock-forming minerals makes accurate
determination of quartz content through chemistry complicated.
Bond work index (kWh t-1)
25.020.015.010.05.00
f
-.50
-1.00
-1.50
-2.00
-2.50
-3.00
-3.50
Fig. 15. Slope of the product size–energy relationship as a function of Bond Work
Index (closing screen size = 355
l
m).
110 K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111
Author's personal copy
Typically, point counting, quantitative X-Ray Diffractometry or
quantitative automated mineralogy (i.e. QEMSCAN or MLA analy-
sis) would be required. However, provided the rock is not too
fine-grained, quartz is easily recognised in lump material and an
empirical visual assessment (±5%) may suffice. A similar situation
applies for porosity. Measurement of porosity by QEMSCAN can
be considered qualitative to semi-quantitative in nature and accu-
rate determination requires mercury porosimetry, which is a costly
procedure. Initial results suggest porosity levels <1.00% are of little
consideration to high voltage breakage.
A high comminution energy input after high voltage treatment
means it is more likely that weakening will off-set the additional
high voltage energy input to result in a net reduction in overall en-
ergy demand. Furthermore, the correlation between high voltage
breakage and Bond Work Index is limited (see Fig. 15).Therefore,
processing circuits treating harder materials or grinding material
to a comparatively small passing size should offer more scope for
potential energy saving after high voltage-induced weakening.
Combining this consideration with the effect of feed size on high
voltage breakage, it is suggested that high voltage breakage is best
implemented in a circuit processing an ore with a high Work Index,
treating relatively coarse (>20 mm) material (i.e. pre-SAG mill, or
possibly pre-ball mill if the feed is coarse enough). The top feed
size for high voltage breakage depends on the electrode geometry,
and the top-end of the feed size effect. It is also suggested finer
material (<10 mm) is removed as these feed sizes may consume
part of the spark energy without undergoing significant weaken-
ing. These suggestion are based on the feed size effect and the
assumption that earlier implementation of high voltage breakage
in a process offers more scope for energy reduction through weak-
ening, and need further investigation to ascertain where potential
benefits from high voltage breakage can be realised most fully.
5. Conclusions
The purpose of this research was to establish what influence
equipment settings, feed size and rock properties have on fragmen-
tation behaviour during high voltage breakage. Key conclusions
drawn from the data and considerations presented in this paper
are:
Rocks being fragmented using high-voltage breakage equip-
ment all experience an initial phase at low energy inputs
(<5 kW h t
1
) during which little size reduction occurs, followed
by a strong decrease in size levelling off towards high energy
inputs.
Total applied energy is the main variable to be considered for
product size (controlled through both number of discharges
and voltage).
The applied voltage controls the amount of energy deposited
per discharge. For the majority of rock types the influence of
voltage on the product size does not deviate from the general
energy-product size relationship for a rock, but some rocks dis-
play a ‘threshold voltage’ below which fragmentation is less
effective.
Voltage gradient between electrodes can be influenced through
voltage and electrode gap. A minimum voltage gradient
(approximately 7 kV mm
1
) needs to exist to reliably achieve
discharges.
Pulse rate can be increased to improve the probability of a dis-
charge occurring.
There is a strong feed size effect, with coarse particles being
considerably more amenable to high voltage breakage than fine
particles.
The amount of particles in the feed-size fraction after treatment
decreases exponentially with energy input. The rate of decrease
can be correlated to tensile strength, suggesting fragmentation
occurs in a tensile stress regime.
Acoustic impedance shows a significant correlation to product
size, providing experimental evidence for shockwaves playing
an important role in high voltage breakage.
Porosity and quartz content are two other rock properties that
can be linked to breakage behaviour. Bond Work Index does
not show a robust correlation to high voltage breakage, but
may be used as a first-order indication of ease of breakage.
High voltage breakage would be best implemented prior to
grinding.
Acknowledgements
We are grateful to the European Social Fund (European Union)
for funding this PhD research and SELFRAG AG for their coopera-
tion and substantial input into the research essential for this paper.
The IOM
3
is gratefully acknowledged for awarding the Tom Sea-
man Travelling Scholarship that made extended stays in Switzer-
land possible. Thanks also go to Dr. Ted Bearman for his advice
and to Mr. J. Flitcroft and Wardell Armstrong and various quarries
around the UK for kindly providing ores and aggregates to perform
research on.
References
Andres, U., 2010. Development and prospects of mineral liberation by electrical
pulses. International Journal of Mineral Processing 97 (1–4), 31–38.
Andres, U., Timoshkin, I., Jirestig, J., Stallknecht, H., 2001a. Liberation of
valuable inclusions in ores and slags by electrical pulses. Powder Technology
114, 40–50.
Andres, U., Timoshkin, I., Soloviev, M., 2001b. Energy consumption and liberation of
minerals in explosive electrical breakdown of ores. Transactions of the Institute
of Mining and Metallurgy C: Mineral Processing and Extractive Metallurgy 110,
149–157.
Armitage, P.J., Worden, R.H., Faulkner, D.R., Aplin, A.C., Butcher, A.R., Iliffe, J., 2010.
Diagenetic and sedimentary controls on porosity in lower carboniferous fine-
grained lithologies, krechba field, algeria: a petrological study of a caprock to a
carbon capture site. Marine and Petroleum Geology 27 (7), 1395–1410.
Bluhm, H., Frey, W., Giese, H., Hoppé, P., Schultheiss, C., Sträßner, R., 2000.
Application of pulses HV discharges to material fragmentation and recycling.
IEEE Transactions on Dielectrics and Electrical Insulation 7 (5), 625–636.
International Society for Rock Mechanics. 1981. In: Brown, E.T. (Ed.). Rock
Characterization Testing and Monitoring: ISRM Suggested Methods. Oxford,
Pergamon.
Deister, R.J., 1987. How to determine the bond work index using lab ball mill
grindability tests. Engineering and Mining Journal (February), 42–45.
Lastra, R., Cabri, L.J., Weiblen, P., 2003. Comparative liberation study by image
analysis of Merensky reef samples comminuted by electric-pulse
disaggregation and by conventional crusher. In: Lorenzen, L., et al. (Eds.),
Proc. XXII International Mineral Processing Congress, vol. 1, pp. 251–260.
Norgate, T., Jahanshahi, S., 2011. Reducing the greenhouse gas footprint of primary
metal production: where should the focus be? Minerals Engineering 24 (14),
1563–1570.
Pirrie, D., Butcher, A.R., Power, M.R., Gottlieb, P., Miller, G.L., 2004. Rapid
quantitative mineral and phase analysis using automated scanning electron
microscopy (QEMSCAN
Ò
); potential applications in forensic geoscience. In: Pye,
K., Croft, D.J. (Eds.). Forensic Geoscience, Pinciples Techniques and Applications,
vol. 232. Geological Society Special Publication, London. pp. 23–136.
Rollinson, G., Andersen, J.C., Stickland, R.J., Boni, M., Fairhurst, R., 2011.
Characterisation of non-sulphide zinc deposits using QEMSCAN
Ò
. Minerals
Engineering 24 (8), 778–787.
Rudashevsky, N.S., Burakov, B. E., Lupal, S.D., Thalhammer, O., 1995. Liberation of
accessory minerals from various rock types by electric-pulse disintegration-
method and application. Transactions of the Institution of Mining and
Metallurgy. Section C: Mineral Processing and Extractive Metallurgy, 104.
Tromans, D., 2008. Mineral comminution: energy efficiency considerations.
Minerals Engineering 21 (8), 613–620.
Wang, E., Shi, F., Manlapig, E., 2011. Pre-weakening of mineral ores by high voltage
pulses. Minerals Engineering 24 (5), 455–462.
Wang, E., Shi, F., Manlapig, E., 2012. Mineral liberation by high voltage pulses and
conventional comminution with same specific energy levels. Minerals
Engineering 27–28, 28–36.
K.P. van der Wielen et al. / Minerals Engineering 46-47 (2013) 100–111 111
