Vertical Roller Mills used for the comminution of mineral

ores
*Adam Mc Elroy1, Dr. Caroline Woywadt2, Bruno Kovacs2

1
Boliden Mineral AB
Finnforsvägen 4
SE-936 81 Boliden, Sweden
(*Corresponding author: [email protected])

2
Gebr. Pfeiffer SE
Barbarossastr. 50-54
67655 Kaiserslautern, Germany

Abstract
Vertical roller mill (VRM) is an established technology in cement and slag grinding applications and Boliden
and Gebr. Pfeiffer have been collaborating on the application of this technology in the minerals processing
industry. In such an application a VRM would function as the milling stage in the primary grinding circuit,
in the role typically filled various combinations of autogenous (AG) mills, semi-autogenous (SAG) mills,
high pressure grinding rolls (HPGR) and pebble or ball mills. The low efficiency of tumbling mills is
documented in literature with Wills (2016) referencing sources which estimate values between 1% and
26% (1), and Bouchard et al. (2019) estimating 9% energy efficiency on average (2). It is anticipated that
VRM can offer improved comminution energy efficiency over tumbling mills.
With regard to water consumption in mineral processing a dry comminution process allows greater
control of water use. The addition of process water would be steered solely by processing downstream of
the comminution circuit, flotation or leaching for example, and not by milling and classification
requirements. Whether this enables a higher solids content in flotation and leaching was not investigated
as part of this paper. The increasing importance of environmental, social and governance (ESG)
considerations strongly motivate the development of more energy and water efficient mineral processing
circuits.
With a view to developing VRM in minerals application three Boliden ores were tested in the Pfeiffer pilot
VRM and supplemented with laboratory scale AG milling characterisation, flotation and leaching test
work. A copper ore, a gold ore, and a complex sulphide ore were chosen for testing.
VRM operating principle
Vertical roller mills are traditionally air-swept mills. A VRM applies a compressive force using fixed rollers
pressing downwards on a rotating grinding table. The material to be ground is fed into the centre of the
grinding table via a feed chute. From there it forms a material bed which passes under the rollers and
undergoes breakage and thereafter flows over the edge of the grinding table. A rising air flow transports
the table overflow upwards to the classifier where finished product continues upwards and out of the mill
while coarse particles flow downwards and are returned to the mill for further grinding. Very coarse
particles which are too heavy to be lifted by the air flow exit the bottom of the mill and require mechanical
transport back to the mill feed. After overflowing from the mill, the fines product passes through a filter
baghouse or cyclone plus filter baghouse in order to separate the solids from the air flow.
The general operating principle is shown in Figure 1 which also shows an alternative set-up to the single
product system, where the classifier coarse is extracted as a product rather than returning to the mill. This
enables a two-product system with the fines extracted via the classifier overflow and the bulk of the
product via the underflow using the so called grits extractor. The achievable product size from a VRM is
in the range between P80 of 1 mm down to P99 of 20 μm.
Comminution is achieved by compressive and shear stress on the particle bed; the surface contact
between the particles distributes the stress in the grinding bed. With application of sufficient stress, micro
cracks are generated. In the grinding bed there are multiple contact points where pressure is applied on
the particles. The compressive comminution with an adjustable proportion of shear forces results in an
energy-efficient size reduction compared to tumbling mills.

Figure 1: Vertical roller mill operating principle (left: standard configuration, right: configuration with grit extraction)

Pilot plant description and standard testing methods

The grindability of a particular mill feed forms the basis for the design of an industrial grinding plant.
Reliable measurement of the properties of the mill feed covered by the term “grindability” is therefore
important. Extensive experience has shown that it is advisable to determine the grindability in a semi-
industrial grinding plant in which the conditions of the future operational plant can be simulated as
accurately as possible.
The Pfeiffer pilot mill used in this study is a semi-industrial vertical roller mill of type MVR with a table
diameter of 400 mm. The mill can be equipped with different classifiers for specific applications. The
process layout of the plant (Figure 2) corresponds to that of an industrial plant.

Figure 2: Process layout of the semi-industrial pilot plant

Different grinding modes can be realized with the MVR pilot mill: “straight grinding” where one finished
product at a certain grain size is generated. When using the pilot plant in “de-sliming” or “selective
grinding” mode, a special classifier with grit extraction is used as shown in Figure 3.
Figure 3: pilot plant with MVR 400, left: standard configuration, right: classifier with grit extraction

The material delivered for pilot testing was pre-crushed in a jaw crusher to a size of 0-8 mm. Standard
procedures (SOPs) are in place for all well-known applications. Those SOPs include parameters such as
specific grinding force, grinding parts geometry, air flow suitable for the specific material and required
grind size. Each test is conducted for a grind size range and evaluates the grinding behaviour at different
grind sizes, the throughput of the pilot plant in kg/h at each grind size, and the absorbed power of the mill
main drive. The limits of unstable operation are investigated during the tests as well and are considered
for the industrial mill rating.
Samples of the VRM product were taken and used for rougher flotation test work to evaluate the
metallurgical response with this milling technology. A Wemco flotation cell was used. The same cell was
used for tests on the VRM product in Kaiserslautern and on rod mill products in the Boliden laboratory,
with the cell being shipped between laboratories as necessary during different phases of testing.
The Kankberg sample also underwent leaching tests as the processing flowsheet for this ore incorporates
both flotation and leaching for gold recovery. Several grind sizes were tested for each ore, with and
without the grits extractor operating.
For each of the ores tested at the Pfeiffer pilot facility, a sub-sample of the pilot test sample was taken for
and sent to the Boliden research laboratory for laboratory scale autogenous grinding testing. This was
used to estimate the energy consumption in an autogenous milling circuit and benchmark the energy
consumption in a VRM circuit determined by the pilot testing.

Boliden Test Samples

Aitik
Located in northern Sweden, Aitik is a low-grade porphyry copper ore. Copper is present as chalcopyrite,
pyrite and pyrrhotite are the main sulphide gangue mineral. Other significant minerals include magnetite,
molybdenite and sulfates (3). Non-sulphide gangue is composed mainly of quartz, feldspar and mica. Aitik
is an open pit mine producing approximately 45 Mt/yr ore.
The Aitik comminution circuit is fully autogenous with 22.5 MW AG mills and 10 MW pebbles mills. Pebble
crushers are also available for crushing of excess pebbles and AG mill scats. The mills operate in closed
circuit with screw classifiers.
Chalcopyrite is recovered via flotation in a rougher-scavenger-cleaner configuration.

Kankberg
The Kankberg Mine is located in the eastern part of the Skellefte mining field (4). Kankberg is a gold
telluride deposit hosted mainly by quartz-feldspar porphyry, volcaniclastic and quartz-andalusite rock
types (4). Kankberg is an underground mine.
The Kanberg ore is campaign treated at the Boliden area concentrator alongside ores from other mines.
The Boliden area concentrator processes approximately 1.8 Mt/yr ore and 200 kt/yr slag. The Boliden
Area concentrator has three processing lines, A line for slag, B and C lines for ore. The ore processing lines
are fully autogenous with AG and pebble mills operating in closed circuit with hydrocyclones. Depending
on the ore type, milling and classification can be supplemented by centrifugal gravity separation and flash
flotation.
Gold is recovered via flotation and leaching, with the flotation concentrate undergoing a warm leaching
process and flotation tails leaching at ambient conditions.

Garpenberg
The Garpenberg mine is located in the Bergslagen district of south-central Sweden. Garpenberg consists
of multiple polymetallic deposits containing zinc, lead, copper, silver and gold as payable metals. The
mineralization consists mainly of pyrite, sphalerite, galena and silver minerals (5). The minor amounts of
copper are present as chalcopyrite. Garpenberg is an underground mine producing approximately 3.2
Mt/yr ore.
The fully autogenous Garpenberg comminution circuit consists of a 5 MW AG mill and a 2.5 MW pebble
mill. The mills operate in closed circuit with hydrocyclones. Milling and classification is supplemented by
centrifugal gravity separation and flash flotation.
The base metals are recovered via flotation first with a copper-lead bulk float, with cleaning and copper
lead separation. Zinc is then floated from the copper lead tails in a sphalerite flotation circuit.

Test work results and benchmarking

In comparison to VRM, AG and pebble mills represent a larger share of total comminution circuit energy.
Therefore, it is important to compare total circuit specific energy rather than just milling in isolation. The
main auxiliary equipment in autogenous mill circuits consists typically of mill feed conveyor belts with
associated feeders, and classifier feed pumps. Other minor electrical consumers include centrifugal
concentrators, screens, and flash flotation cells.
In a VRM circuit the fan, classifier, and hot gas generator (if required) are expected to represent a larger
portion of the total circuit energy consumption than the auxiliaries in an AG mill circuit.
The energy consumption of auxiliary equipment in the AG circuits is based on actual plant data for Aitik
and Garpenberg, Kankberg auxiliaries were roughly estimated based on the Garpenberg data. Average
data, corrected for throughput was collected to provide a basis for circuit auxiliaries. AG circuit mill energy
consumption is based on laboratory scale characterisation of a sub-sample of the material sent to Pfeiffer
for VRM testing. Boliden’s standard AG test was used.
VRM mill energy and total circuit energy was calculated based on the pilot tests.

Comminution results
Aitik
As Aitik comminution circuit consists only of mills in closed circuit with classifiers (when the cone crushers
are being bypassed) it is relatively straightforward to compare to a VRM circuit which is also milling in
closed circuit with a classifier. Mill and overall circuit energy consumption for both AG and VRM
technologies are summarised in Table 1. Moisture levels were not investigated as part of the Aitik pilot
test.

Table 1. Aitik mill and mill circuit specific energy consumption for AG and VRM circuits.

AG AG VRM VRM VRM VRM/AG

P80 mill circuit mill Drying circuit circuit
w/o w/o
drying drying
µm kWh/t kWh/t kWh/t kWh/t kWh/t %
154 13,6 15,2 3,2 - 7,5 49
180 12,4 13,9 2,4 - 6,3 45
233 9,9 11,1 2,0 - 5,6 50

Kankberg
Kankberg comminution results for the typical plant p80 are summarised in Table 2. The sample is harder
than average for Kankberg though within normal variations. Up to 2% moisture it is expected that the
VRM could operate without drying. The exact point at which feed moisture becomes problematic for
Kankberg and requires drying by using a hot gas generator requires continued investigation.

Table 2. Kankberg mill and mill circuit specific energy consumption for AG and VRM.

Feed AG AG VRM VRM VRM VRM/AG

P80 Moisture mill circuit mill Drying circuit circuit
µm % kWh/t kWh/t kWh/t kWh/t kWh/t %
45 2,0 37,6 40,6 9,4 - 21,6 53

Garpenberg
Results of the Garpenberg comminution test work are shown in Table 3. For mechanical energy only the
VRM shows significant energy savings for the both the mill in isolation and the total circuit energy.
Similarly to Kankberg, with Garpenberg ore it is expected that up to 2% feed moisture is tolerable in a
VRM. With the exact tipping point for drying energy to be determined. However, if the threshold is crossed
where the moisture levels require drying energy via a hot gas generator then the VRM circuits goes from
consuming 79% energy compared to the AG circuit to 288-345% at 3.5%-4.5% moisture level. These
moisture levels have been used to provide an indicative example. Further investigation is required to
determine what levels would be manageable in a full-scale mill. The specific heat capacity of the ore has
been assumed until a more accurate value can be obtained, adding some uncertainty to the calculated
values. The Garpenberg comminution circuit includes centrifugal gravity separation and flash flotation,
which cannot be included in a VRM circuit.

Table 3. Garpenberg mill and circuit specific energy consumption for AG and VRM (without drying) circuits.

Feed AG AG VRM VRM VRM VRM/AG VRM VRM/AG

P80 Moisture mill circuit mill Drying circuit circuit circuit circuit
w/o w/o with with
drying drying drying drying
µm % kWh/t kWh/t kWh/t kWh/t kWh/t % kWh/t %
115 2,0 16,4 19,7 5,6 - 15,5 79 - -
115 3,5 16,4 19,7 6,4 42,7 14,0 71 56,7 288
115 4,5 16,4 19,7 5,6 54 14,0 71 68,0 345

Process results
Aitik
Of note with the flotation results on the Aitik sample is that all results are well below typical plant results.
However, the laboratory tests in this study, VRM versus tumbling mill, can still be compared to each other
even if neither case compares well to typical full-scale results.
Flotation samples from ranging from p80 154 µm to 237 µm were tested for flotation response, with the
typical plant p80 being 190 µm in recent operation, slightly higher than the design nominal of 180 µm.
VRM in straight grinding mode outperformed the reference tests using a rod mill by a wide margin.

Kankberg
Despite being somewhat low grade by Kankberg standards, both grinding methods achieved recoveries
within the normal range, though with a clear difference between VRM and rod mill. Flotation rougher
concentrate gold grade and recovery were improved using a VRM.
Concentrate leaching results to date show no significant recovery difference between the grinding
methods in this leaching step. However, as warm leaching has a higher recovery than the ambient leaching
the increased amount of gold reporting to warm leaching is expected to result in an overall improvement
in overall gold recovery.
Tailings leaching test work results are pending and required to complete an overall balance.

Garpenberg
Evaluation of the Garpenberg process results is ongoing.
Development challenges
Bench scale testing
As discussed above in section ‘Pilot plant description and standard testing methods’, testing in the semi-
industrial scale pilot mill is recommended for definitive, final sizing of mills. While this level of accuracy
and reliability is of course required to proceed with a mill installation, it is not required for early-stage
characterisation work during scoping study or prefeasibility study level for instance. Green field ores
especially would benefit from the availability of laboratory scale test work.
The tests discussed in this paper were conducted with 10 t samples. Though the goal of the test work was
not a standard characterisation test but to investigate processing characteristics at different mill
conditions and product grind size. A typical characterisation test in the VRM pilot facility requires 1-2t.
While operating mines can readily provide samples currently processed ore in volumes of this magnitude,
future ores and greenfield sites may only have drill core available. Particularly at scoping study level, there
can be cases where only tens of kilograms are made available for metallurgical testing. Larger composite
samples can be obtained at later stages of study, pre-feasibility and feasibility level. Though for variability
testing Wills and Finch (2016) suggest that the:
“Preliminary stage of a large project may need 35 samples to demonstrate the variability
of the ore body and allow estimate of equipment size (albeit imprecise)”.
Such a number of variability test could prohibit a 1-2 t per test sample requirement. 35 discrete 1-2 t drill
core samples available for comminution testing is a challenging requirement.
A summary of established comminution tests in the industry and internal Boliden test methods is listed in
Table 4. There are several tests in the 10-20 kg sample requirement range. The emergence of the
Geopyöra test with a circa 2 kg sample requirement (6) allows calculation of Axb and Bond indices with
low sample amounts.
A laboratory scale VRM characterisation test with sample quantity in the range of the tests listed in Table
4 would enable the inclusion of the technology in early study stages where test work is limited to small
amounts of drill core. Most current methods employ impact or abrasion methods, neither of which are
the dominant breakage mechanism in a VRM. The Geopyörä test applies a crushing action which is more
similar to the compressive breakage in a VRM, this combined with the low sample quantity required make
it the most relevant of the existing comminution characterisation tests for use in VRM studies.
A purpose designed laboratory scale VRM test is also an option that could be developed.
 Table 4. Laboratory tests for comminution circuit design (7), (8), (9), (10), (6), (11), (12).

Test Use References Sample Material size Minimum

quantity, core size
kg
Bond impact Conventional crushers Bond 1946, 1961 40-50 2-3 in. PQ
crushing test
SAG mills Barrat and Allan 1986 - -
Bond rod mill Rod mills/Ball mills Bond and Maxon 10-15 100% - 12.7 NQ
work index 1943;Bond 1961; Rowland mm
1982
SAG mills Barrat and Allan 1986 - -
Bond ball mill Ball mills Bond and Maxon 1943; 15 100% - 3.35 NQ
work index Bond 1961; Rowland 1982 mm

SAG mills Barrat and Allan 1986 - -

Bond abrasion Wear prediction for Bond 1963; Giblett and 2 12-19 mm NQ
index crushers and tumbling mills Seidell 2011

SMC test Conventional crushers, Morrell 2004a-c, 2009 20 19-31 mm NQ

AG/SAG mills, HPGRs

JK drop weight SAG mills, conventional Napier-Munn et al. 1996 75 13.2-63 mm PQ

test crushers

SPI test SAG mills Starkey et al. 1994 10 80%-12.7 NQ

mm
SAGDesign test SAG mills Starkey and Larbi 2012 15 80%-19 mm NQ

Boliden fall test AG mills Boliden procedure 100 200 mm Ore sample

Boliden abrasion AG/pebble mills Boliden procedure 20 5-60 mm Ore sample

test
15 5-30 mm Crushed half
core
Geopyörä Conventional crushers, Chavez et al. 2020; Bueno 1.5-3 13.2-31.5 BQ
AG/SAG mills, ball mills, et al. 2021; Bueno et al. mm
HPGR and rock mechanics 2023; Bueno et al. 2023

Moisture limits
In the cement and slag grinding industries the VRM is operated as a fully dry process, with the hot gas
generator providing the drying energy required to remove feed moisture during the milling and
classification process and produce a fully dry product. The classifier and bag filter used as standard in VRM
circuits are therefore not designed to tolerate the levels of moisture that can be found in ores, listed here:
• Dunne (2016) “New feed containing typically 2-3% and up to 10% moisture” (7).
 • Rodríguez et al (2023) – 2-5% (13).
• Bleiwas (2012) – 2-5% (14).

Various Boliden studies have used values between 1-5% moisture in the mill feed. However as autogenous
milling circuits are not sensitive to moisture in the same way that VRM circuits are this figure has not been
examined particularly closely. Water is added to the mill feed chute diluting the feed to 65-70% solids, or
30-35% moisture content.
The milling action itself does produce some heat, and in hard fine-grained ores this effect will be larger
than softer coarser ores. In some cases this may be sufficient to dry the material, not to zero percent
moisture, but to a level that the classifier and filter can tolerate. The Kankberg pilot test for example,
successfully tested the circuit with 2.0% feed moisture without actively heating the process air with the
hot gas generator. At 2.8% moisture and 4 L/h water injection in the mill low heating was required.
With flotation and leaching being the most common processing methods, both wet processes, consuming
energy to dry ore only to subsequently add water is wasteful. Table 3 shows the extent of this energy
consumption, the additional CO2 emissions should also be considered. If the mill itself can function
acceptably with the moisture contents in run of mine ore development of classification and solid-air
separation that could handle the same moisture content is worth attempting.

Circuit design
Greenfield
A1 Straight grinding with a single VRM product to downstream processing.
This is the simplest VRM circuit and can could be operated with or without storage between milling and
downstream processing. A flowsheet for this case is shown in Figure 4. A silo or silos between the filter
and flotation feed are necessary for stable operation and feed to the flotation plant. However, the sizing
of the silo(s) could be chosen either for the minimum residence time for stable operation of both milling
and processing circuits or large enough to fully decouple the milling and processing circuits. A 12 or 24
hour milled ore storage would allow short maintenance or inspection stops of the mill circuit without
disrupting downstream processing. This would also allow a very controlled and stable feed to the
processing circuit. The risk with this approach is with flotation of sulphide ores which are sensitive to
oxidation. Test work with Boliden ores has not demonstrated degradation of ore samples that were
floated as soon as possible after milling. How susceptible sulphide ores are to oxidation and how soon
after VRM milling they need to be floated to avoid negative oxidation effects should be investigated on a
case by case basis. Oxide ores do not have this vulnerability and decoupling milling and processing could
be advantageous for these ores.
Figure 4. Straight grinding

A2 With the grits extractor operating and two products from the VRM circuit
This circuit arrangement produces two products from the VRM. A coarse fraction extracted via a screw
feeder collecting the classifier coarse and a fines product flowing out of the classifier with the process air.
This flowsheet is illustrated in Figure 5.

Figure 5. VRM circuit with grits extraction

A3 Coarse ore sorting between secondary crushing and VRM followed by dry pre-
concentration of the VRM product(s).
This flotation aims to minimise process water consumption and the production of wet tailings option and
is illustrated in Figure 6. Variations on this flowsheet can also be considered.
• Coarse ore-sorting only, no dry fines pre-concentration.
• Dry fines pre-separation, no coarse ore-sorting.
• With the grits extraction operating and separate treatment of VRM grits and fines.

Separating part of the waste rock in dry sorting and separation processes will reduce overall water
consumption as no process water will be entrained in these tailings products.
Boliden and Luleå University of Technology are currently investigating the potential of dry separation
after dry grinding as a pre-concentration method in advance of flotation.

Figure 6. Coarse ore sorting and fine pre-concentration.

Brownfield integration
Incorporating a VRM in a brownfield tumbling mill circuit is an option when production increases are
required. For example, in pebble crushing duty. A screened pebble fraction even if has been extracted
from an AG/SAG mill should be low in moisture due to the lack of fines which could entrain water. Plant
surveys in Boliden’s Aitik and Kevitsa comminution circuits indicate moisture content of ca. 1%. This is
below the expected moisture level where drying energy would be incurred
Summary and conclusion
VRM compares quite favourably to AG mills, both in terms of comminution energy and metallurgical
performance when feed moisture is within tolerable limits.
If drying energy is required then this can lead to a VRM circuit being significantly less effective than the
AG mill circuits used as a benchmark in this study.
Several mill circuit flowsheets have been presented for consideration, and further flowsheet variations
are possible. The use of dry comminution provides much potential for new circuit options.
References
1. Wills, Barry A and Finch, James A. Will's Mineral Processing Technology, An Introduction to the Practical
Aspects of Ore Treatment and Mineral Recovery. 8th. Oxford : Elsevier Ltd., 2016. ISBN 978-0-08-097053-
0.
2. Breaking Down Energy Consumption in Industry Grinding Mills. Bouchard, J, et al. 4, s.l. : CIM Journal,
2019, Vol. 10, pp. 157-164.
3. Boliden Mineral AB. Aitik PERC Technical Report 2021. Boliden : s.n., 2021.
4. —. Resources and Reserves Kankberg 2019. Boliden : s.n., 2019.
5. —. Garpenberg PERC Technical Report. Boliden : s.n., 2023.
6. The Double Wheel Breakage Test. Bueno, M, et al. s.l. : Minerals Engineering, 2021.
7. Dunne, Robert C, [ed.]. SME Mineral Processing & Extractive Metallurgy Handbook. Englewood :
Society for Mining, Metallurgy & Exploration, 2019. ISBN 978-0-87335-385-4.
8. Boliden Mineral AB. Metodbesrkivning Falltest. s.l. : Boliden Mineral AB.
9. —. Metodbeskrivning Slitagetest. s.l. : Boliden Mineral AB.
10. Applied Geometallurgy at Agnico Eagle’s Kittila Operation using the Geopyörä Breakage Test. Bueno,
M, et al. Vancouver : SAG Conference, 2023.
11. Extensive Validation of a New Rock Breakage Test. Bueno, M, de Almeida, T and Powell, M.
Comminution and Comminution Circuits Optimisation, Volume II, s.l. : Minerals.
https://doi.org/10.3390/min13121506 .
12. The Geopyörä breakage test for geometallurgy. Chavez, M, et al. s.l. : IMPC, 2020.
13. Water footprint for mining process: A proposed method to improve water management in mining
operations. Rodríguez, Jorge E, Razo, Israel and Lázaro, Isabel. s.l. : Elsevier Ltd., 2023, Cleaner and
Responsible Consumption, Vol. 8. ISSN 2666-7843.
14. Bleiwas, Donald I. Estimated Water Requirements for the Conventional Flotation of Copper Ores.
Reston : US Geological Survey, 2012.