Int. J. Miner. Process.

74S (2004) S65 – S74

www.elsevier.com/locate/ijminpro

Dry fine comminution in a stirred media mill—MaxxMillR

Yanmin Wanga,c,*, Eric Forssberga, Jens Sachwehb
a
Department of Chemical & Metallurgical Engineering, Luleå University of Technology, S-971 87 Luleå, Sweden
b
Maschinenfabrik Gustav Eirich GmbH & Co KG, D-747 32 Hardheim, Germany
c
Institute of Material Science & Engineering, South China University of Technology, 510640, Guangzhou, China

Abstract

A large-scale continuous stirred media mill called the MaxxMillR system in dry mode was used to investigate the
comminution characteristics. A limestone powder below 2 mm was used as a feed material. A statistical experimental design
was prepared with the major influencing parameters of the MaxxMillR, such as rotational speed of stirrer, feed rate and bead
size. The fineness of the ground products obtained under various conditions of the MaxxMillR and the corresponding energy
utilisation have been evaluated and analysed. The results have been empirically correlated with the operating parameters. In
addition, product size–energy input relations can be described empirically, independent of the parameters used.
D 2004 Elsevier B.V. All rights reserved.

Keywords: comminution; stirred media mill; limestone; powder; fine particle

1. Introduction is below 200 Am. In other words, the designs of

these machines do not allow an efficient comminu-
The comminution devices used for production of tion for the coarse particles in the feed material.
micron and submicron particles are mainly high- A stirred media mill named MaxxMillR, which
speed stirred media mills. These mills offer signifi- has been developed recently by Maschinenfabrik
cant reductions in specific energy consumption Gustav Eirich GmbH, Germany, expands the appli-
compared to conventional tumbling mills. The cation of stirred media mills to a feed size of several
reductions are mainly due to the use of small millimetres (Sachweh, 1998). This mill is a comple-
grinding media with high stress intensities in the ment with respect to the existing stirred media mills.
mills (Becker et al., 2001). These mills provide an It can perform a fine dry/wet comminution for coarse
effective size reduction/surface enhancement in the particles. The MaxxMillR is a new version of the
product. However, the required feed size for the stirred media mill with its unique design and
materials suitable for the existing stirred media mills operating principle.
This paper concerns the evaluation of the per-
* Corresponding author. formance of the dry MaxxMillR system grinding a
E-mail address: Wang.Yanmin@ltu.se (Y. Wang). limestone material below 2 mm by a statistical
0301-7516/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.minpro.2004.07.010
S66 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74

approach. In the experiments, the major parameters eccentricity. The grinding media, like steel, glass or
(rotational speed of stirrer, bead size and feed rate) ceramic beads with sizes of 3 to 10 mm, are filled up to
of the dry system have been statistically analysed. In 80 vol.%. A coarse feed material enters the machine
addition, product size–energy relations obtained in through the pipe in the wall scraper together with the
the system have been also discussed. carrier fluid (air or water). The material will be mixed
intensively with the beads, which have a pulverising
effect on the product. The fine product will be sucked
2. Experimental from the upper layer of the beads through the product
outlet. Most of the grinding beads do not leave the mill
2.1. Materials used because of their weight. The coarse material and a few
small beads will be recirculated by an air classifier into
Limestone material b2 mm (d 50: 890 Am; specific the bottom of the grinding chamber with the feed
surface area: 0.26 m2 g
1) from Partek Nordkalk OY, material. However, for high-viscosity slurries, a sieve
Finland, was selected as an experimental sample. X- is needed to retain the beads. The MaxxMillR is most
ray diffraction analysis identified that calcite (CaCO3) effective with a feed particle size up to 5 mm. The
is a common mineral in the sample. Berolamine-10 or throughput ranges from 100 kg/h to 20 t/h depending
BA-10 (AKZO NOBEL Chemicals GmbH, Germany) on the mill size and the product requirement. This mill
was used as a dispersant for dry comminution of the can be used for milling of the hard components in the
powdered limestone material in the MaxxMillR. ceramic industry and for grinding of minerals like
limestone and quartz in the mineral industry.
2.2. MaxxMillR system An experimental system with the MaxxMillR for
continuously performing dry grinding tests on a
Fig. 1 shows a general design of the MaxxMillR. production scale was used in this study (see Fig. 2).
This mill consists mainly of a rotating drum, a stirrer A variable screw conveyor, which together with the
and a stationary wall scraper with an integrated filling feed tank, is mounted on load cells. The conveyor’s
pipe. The grinding chamber capacity is 190 l with an delivery rate can be controlled with high precision by
installed power rating of 45 kW. The stirrer is located weighing the complete screw tank material and
eccentrically to the middle of the drum with the varying the conveyor drive frequency accordingly. A

Fig. 1. General design of the MaxxMillR.

 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74 S67

wt.% was pumped into the feeding material for dry

comminution. The material was dry-ground to
produce different products under various conditions
of the mill, such as bead size, rotational speed of
stirrer and feed rate. The ground product from the
upper layer of the grinding beads in the chamber was
discharged by a vacuum extractor without using a
sifter and then collected in a filter. In each experi-
ment, a powdered sample was taken in the product
tank for analysis.

2.4. Measurements

Particle size analysis was carried out using an X-

ray Sedimentometer SediGraph 5000 D (Microme-
retics, USA). The particle size range varied between 1
and 150 Am for the measurement. The particles in the
sample were deagglomerated with ultrasonic waves
and an aqueous solution with a dispersing chemical
Fig. 2. Experimental system with the MaxxMillR for dry grinding. CALGON or Na(PO3)6 at the concentration of 0.1%
prior to the analysis. The specific surface area was
sifter is installed with a follow-up filter for dry
grinding mode. The grinding media used in this study
Table 1
was ceramic beads with two sizes of 4–5 and 5–7 mm. Experimental design for dry grinding coarser limestone powder by
MaxxMillR system, L18 (2132)
2.3. Experimental design and procedure Test no Parameter level Responses
Size of Feed Rotational speed
A statistical experimental design of multiple bead (mm) rate (kg h
1) of stirrer (rpm)
variable with mixed levels (L18 213n ) was employed 1 1 1 1 R1
to perform the dry grinding experiments, as shown in 2 1 1 2 R2
Table 1. The independent variables or parameters, 3 1 1 3 R3
which affect the performance of the MaxxMillR, were 4 1 2 1 R4
selected for investigation as follows: 5 1 2 2 R5
6 1 2 3 R6
7 1 3 1 R7
– Size of bead (mm), 8 1 3 2 R8
– Rotational speed of stirrer (rpm), and 9 1 3 3 R9
– Feed rate (kg h
1). 10 2 1 1 R10
11 2 1 2 R11
12 2 1 3 R12
The responses or dependent variables for the 13 2 2 1 R13
statistical analysis/evaluation of the experimental 14 2 2 2 R14
results are energy utilisation [new BET surface area/ 15 2 2 3 R15
energy input (DS/E m)] and the median size of the 16 2 3 1 R16
ground product (d 50). 17 2 3 2 R17
18 2 3 3 R18
The MaxxMillR system works continuously in Level 1 2 3
dry mode. The powdered material in the hopper at Size of bead (mm) 5–7 4–5
the top of the system was fed into the grinding Feed rate (kg h
1) 300 500 700
chamber through a feed pipe near the bottom of the Rotational speed 350 480 588
chamber. Berolamine-10 diluted at an amount of 0.1 of stirrer (rpm)
S68 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74

measured by a BET method with the Flow Sorb II Table 2

2300 (Micromeretics). A representative amount was Results of ANOVA for dry comminution of the coarse limestone
(b2 mm) by MaxxMillR system
taken from each sample and was let to dry in an oven
Source Sum of df Mean F-ratio Probability
at 353 8K (or 80 8C) in order to remove the residual
squares square
moisture of the sample before the measurement. X-ray
(a) Energy utilisation (DS/E m)
diffraction (XRD) was made with a Siemens D5000
Main parameter
powder diffractometer using Cu Ka radiation (Sie- B (bead 0.13885 1 0.1388 0.76 0.4068
mens, Germany). In this study, an active AC power size, mm)
(kW) and the energy consumption (kWh) were F (feed 5.39206 2 2.6960 14.70 0.0015
recorded by taking the power factor into account. rate, kg h
1)
R (rotation of 1.38267 2 0.6913 3.77 0.0447
The net energy consumption of the MaxxMillR
stirrer, rpm)
system, the mass specific energy input, E m (kWh
t
1), was determined by E m=( P-P 0)/F, where P is the Covariates
active power draft of rotor, P 0 is the idle power draft B F 0.06163 1 0.0616 0.34 0.5763
without grinding media and ground material, and F is F R 0.03251 1 0.0325 0.18 0.6836
B R 0.35021 1 0.3502 1.91 0.2003
the feed rate of the dry solids.
Residual 1.65053 9 0.1839
Total (corrected) 21.63 17

3. Results and discussion (b) Median size (d 50)

Main parameter
B (bead 0.18809 1 0.1881 0.22 0.6472
3.1. Analysis of variance in statistics
size, mm)
F (feed 6.23543 2 3.1177 3.65 0.0577
Experiments on three major operating parameters rate, kg h
1)
(bead size, feed rate and rotational speed of stirrer) R (rotation of 36.4777 2 18.238 21.37 0.0001
were carried out. A significant difference was stirrer, rpm)
Residual 10.2440 12 0.8536
considered at a confidence level of 95% (ab0.05),
Total (corrected) 53.1452 17
Turky significant differences test (Montgomery,
All F-ratios are based on the residual mean square error.
1991) in the analysis of variance (ANOVA) was
utilised to analyse a difference between two sizes of
beads: a difference among three levels of feed rate probability for the bead size is 40.68%. This proves
and a difference among three levels of stirrer the hypothesis that its variation does not affect the
rotational speed as well. Because the designed results of the energy utilisation. The effect of the
experiments were not replicated, it was assumed bead size has not been demonstrated to affect the
that the three-factor interaction effects would be energy utilisation significantly from a statistical
negligible. The ANOVA was used to examine the viewpoint. This illustrates that a variation of BET
null hypothesis that the variations of each inde- surface area of the ground product appears to be
pendent variable at various levels have no effect on insignificant when coarser limestone below 2 mm
the energy utilisation index (DS/E m). The results of was ground using a bead size range from 4 to 7
the ANOVA in Table 2(a) demonstrate that the mm in the dry MaxxMillR system. In Table 2(a),
probabilities for the null hypothesis that the feed all two-factor covarieties on the DS/E m show lower
rate and the rotation of stirrer have no significant values of F-ratio. All parameters studied do not
effects are 0.0015 and 0.0447, respectively. This interact due to their high significant levels, and
means that the feed rate and the rotational speed of these covarieties are regarded as statistically insig-
stirrer are the primary important variables among nificant effects.
the parameters studied. The variation of each Table 2(b) shows that the rotational speed of the
parameter dramatically influences the energy uti- stirrer is the most dominant effect on the median
lisation (DS/E m) in dry comminution of the coarser size (d 50) of the ground product because its F-ratio
limestone below 2 mm by the MaxxMillR. The is 21.37. The second important parameter becomes
 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74 S69

the feed rate. The probability for this parameter to When introduced into the operating mill, the
have no effect for the null hypothesis is 5.77%. The particles segregate rapidly via a feeding pipe to the
results indicate that the rotational speed of stirrer bottom of the grinding chamber. The beads with the
and the feed rate both are the important factors, particles to be ground circulate radially within the mill
which influence the product fineness obtained in the but can also be removed from the mill top by
dry MaxxMillR system. However, the bead size has virtue of bead motion. The beads (with the
no effect on the variation of the median size of the particles) vacated in this way are replaced and
ground product, as its probability is 64.73%. dispersed by an induced airflow (from vacuum), as
well as radial displacements in the mill. The
particles and beads, which move into the stirrer
4. Effects of parameters core, are subsequently displaced toward the wall of
the grinding chamber.
4.1. Rotation of stirrer and feed rate Lower feed loading in the mill appeared at a low
feed rate, which causes a significant enhancement of
The results of ANOVA illustrate that the per- product fineness (Fig. 5). However, the energy
formance of the dry comminution using the inducing to beads for the breakage of particles may
MaxxMillR is significantly affected by the rotational not be efficiently utilised at a lower feed rate. Clearly,
speed of the stirrer and the feed rate. Fig. 3 shows a decrease in the feed rate results in a widening of the
the product size distributions with respect to the normalised residence time distribution (Stehr and
effect of rotational speed of stirrer at various feed Schwedes, 1983).
rates. As the rotational speed is increased, the Both rolling and sliding and direct impacts occur
curves of the size distribution shift to a finer size, in the stirrer of the MaxxMillR. Rolling at a low
resulting in a reduction of the median size and an rotational speed may occur in the mill volume
increase of the specific surface area (see Fig. 4). For outside the core. In addition, because rolling on the
instance, a rotation of 350 rpm produced a median bottom of the grinding chamber differs from that on
size of 13.6 Am with its BET surface area of 3.07 the mill sides in terms of the forces and velocities
m2 g
1, while at 588 rpm a median size of 8.3 Am characteristic of each. The rolling in the different
with the surface area of 4.03 m2 g
1 was obtained mill regions will have correspondingly different
at the feed rate of 500 kg h
1 for the case of using efficiencies. It is considered that rolling at the
5- to 7-mm bead size. It is also shown that the bottom can affect a plastic deformation of the
product fineness increases with decreasing feed rate particles to be ground there. The weight of the
at given rotational speeds of stirrer and bead sizes. beads provides a normal force on the particles
An increase in the rotational speed of stirrer or a located at the bottom. This may be an advantage
decrease in the feed rate increases the energy input of the MaxxMillR design, which feeds the coarse
per unit mass. particles from the bottom. The elastic limit of the
Rydin et al. (1993) observed the milling dynamics particles is exceeded if the force exceeds the hard-
of an agitated media mill in dry mode. It was found ness of the particle on its contact area. Moreover, the
that the rotation velocity gradients exist along both stresses on the wall of the grinding chamber
radial and vertical directions in a stirrer at various originate from centripetal acceleration of the stirred
rotations. These give rise to sliding/rolling events, beads. It was calculated that the rolling force on the
and occurrences are far more common than are direct wall is smaller than the value for the mill bottom.
impacts, which occur more frequently in the stirrer Thus, it is unlikely that rolling alone can produce the
core. However, direct impacts do occur outside of plastic deformation on the chamber wall. It is also
the core, and although they are relatively few in supposed that impacts of beads are responsible for
numbers, they take place at a rapid rate. The the plastic deformation of particles. The pertinent
centripetal force displaces beads toward the chamber parameters are the impact velocity and frequency, as
wall. This force arising from the mill rotation may be well as the variation of these parameters with the
considered analogous to a grinding energy. grinding chamber.
S70 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74

Fig. 3. Particle size distributions of the ground products obtained at various rotations and feed rates.
 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74 S71

Fig. 4. Median size and specific surface area of the ground products obtained at various rotations of stirrer and feed rates.

4.2. Bead size mill is not related to the bead size range used in this
study when coarse limestone below 2 mm was
According to some studies (Stehr, 1988; Wang and ground. This resultant insignificance may be, because
Forssberg, 1997), it is possible to optimise the grinding a ratio of median size of feed treated (d 50=890 Am) to
efficiency of the stirred media mill if the bead size is mean size of each bead is slightly different for two
chosen to optimally correspond to the feed size. This sizes of beads used, i.e., between 0.197 for the bead
is because the impact events of the small beads are size of 4–5 mm and 0.162 for the bead size of 5–7
insufficient for the coarse particles in the feed, mm. The coarser beads of 5–7 mm can be used as the
resulting in a poor grinding efficiency. The velocities grinding media when coarse limestone is dry-ground
of small beads in the tangential direction are low when in the MaxxMillR.
they are pressed tightly, which leads to a reduction of
their abrasive energies. The large beads provide a
higher velocity gradient difference. However, it is 5. Energy utilisation
interesting to see in Fig. 4 that the variation between
two levels of the bead sizes (4–5 and 5–7 mm) appears Fig. 5 shows the variations of the energy utilisation
to have an insignificant effect on the fineness of the (DS/E m) with respect to the effects of the feed rate and
ground product obtained in the MaxxMillR. This the rotation of stirrer. It is apparent that the energy
means that the performance of dry comminution in the utilisation is significantly affected by either the feed
S72 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74

MaxxMillR in this work. At a lower rotational speed,

it was noticed that the grinding beads tended to roll
over each other and against the wall, producing an
effect like numerous tiny roll crushers operating. At a
higher rotation however, the action in the mill was
much more violent, resulting in a much higher degree
of impact breakage. Because this latter breakage
mechanism is much less efficient than slow compres-
sion or attrition, the gentle action at a lower rotation
may have resulted in a more efficient breakage
mechanism.
Insufficient probabilities for beads grinding the
particles in the mill with a high bead load occur at a
low feed rate below 300 kg h
1. The energy loses in
an inefficient friction/collision of the beads rather than
an efficient comminution of the particles. However, if
the particle charge with the active beads in the mill is
nearly optimum, the energy can be efficiently utilised
for the grinding action. Fig. 5 also shows that an
optimum energy utilisation may occur at 500 kg h
1
when limestone below 2 mm was dry-ground.

Fig. 5. Energy utilisation with respect to the effects of rotation of

stirrer and feed rate.

rate or the rotational speed of stirrer. The energy input

to the MaxxMillR is utilised more efficiently at a
lower rotation of the stirrer and a higher feed rate.
This conclusion seems to agree with the previous
findings on comminution of coal with the stirred bead
mills (Stehr and Herbst, 1984). According to their
explanation, the vortex created in the core of the mill
at a high rotation of the stirrer reduces the effective
volume for comminution, and thus the specific energy
input of the mill is utilised less efficiently. At a lower
rotation, the central vortex is reduced and the effective
mill volume and the total mill volume become nearly
identical. The evidence of a central vortex could be
clearly observed during the experiments. Thus, the
MaxxMillR performs more energy efficiently at a
lower rotational speed of the stirrer. The mode of the Fig. 6. Fineness of the ground products (d 50 or DS) as a function of
breakage may also play a role in the efficiency of the specific energy consumption.
 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74 S73

6. Empirical prediction 7. Conclusions

For the purpose of prediction, empirical models for Experimental results indicate that the main factors
the estimation of the performance of dry grinding a influencing the energy utilisation (DS/E m) in a
coarse limestone material using the MaxxMillR decreasing order of importance are feed rate and
system were established from the results by the rotational speed of stirrer. The rotation of stirrer is the
statistical approach. The variables involved were the most dominant effect on the median size (d 50) of the
bead size (B), the feed rate ( F), and the rotational ground product. The bead sizes in the range of 4 to 7
speed of stirrer (R) as well. No interaction effects were mm did not influence the grinding results.
considered in these models due to their insignificant An increase of rotational speed of stirrer or a
effects. The following estimated models for the decrease of feed rate results in an increased fineness of
energy utilisation (DS/E m) and the median size of the ground product with a higher specific energy
the ground product (d 50) could be thus expressed for input. Energy to the MaxxMillR for dry comminution
dry comminution of the limestone material. of limestone below 2 mm is utilised more efficiently
at a lower rotation of stirrer or a higher feed rate.
DS B0:1226 F 0:4108

¼ k1 Rsq ¼ 76:91% ð1Þ Empirical models can be used to predict the
Em R0:2641 comminution characteristics with respect to the major
parameters in the MaxxMillR in dry mode. Product
B0:0805 F 0:1534

d50 ¼ k2 Rsq ¼ 78:08% ð2Þ size–energy input relations have been also established,
R0:6128 independent of the operating parameters used.
where k 1 and k 2 are constants, equalling to 2.11104
for the DS/E m in Eq. (1), and 1.57102 for the d 50 in
Eq. (2), respectively. It is seen that the empirical Acknowledgements
models may be appropriate for the predication due to
their higher correlation indications. The models are This work was supported by the Swedish Mineral
primarily valid for the ranges of the parameters used Processing Research Association (MinFo). The
for the dry grinding of the limestone in this authors thank Maschinenfabrik Gustav Eirich GmbH
comminution device. & Co KG, Germany for providing the MaxxMillR
The prediction of comminution behaviour in the system, as well as the corresponding experimental
MaxxMillR and the subsequent development of facility, and Partek Nordkalk OY, Finland, for
scale-up design relationships require a mathematical supplying the limestone sample. Dr. Stefan Gerl of
model relating energy input to the resultant product Eirich is thanked for sincere assistance and valuable
fineness. The product median size (d 50) and the new discussions during the tests.
surface area (DS), respectively, are plotted as a
function of specific energy input in Fig. 6. The References
straight lines in the figures were drawn by using the
following equations: Becker, M., Kwade, A., Schwedes, J., 2001. Stress intensity in
stirred media mills and its effect on specific energy requirement.
d50 ¼ 111:5E
0:61 ð3Þ Int. J. Min. Process. 61 (3), 189 – 208.
Montgomery, D., 1991. Design and Analysis of Experiments, 3rd
DS ¼ 0:338E0:57 ð4Þ ed. John Wiley & Sons, New York.
Rydin, R.W., Maurice, D., Courtney, T.H., 1993. Milling dynamics:
with the correlation coefficients of 0.78 and 0.94. The part 1. Attritor dynamics: results of a cinematographic study.
fact that all the data obtained by dry comminution of Metall. Trans. 24A, 175 – 185.
limestone with the MaxxMillR are relatively well- Sachweh, J., 1998. Agitated media mills for dry and wet grinding
described by a single straight line is important for applications. Keram. Z. 5, 3 – 11.
Stehr, N., 1988. Mfglichkeiten Zur Verhinderung der Machkör-
scale-up. It is possible to predict the fineness of the perverpressung in kontinerlich betriebenen Ruhrwerksku-
ground product (d 50 or DS) for a given energy input, gellmuehlen, Lecture, Jahrestreffens der Verfahrensingenieure,
independent of the operating parameters used. Hannover.
S74 Y. Wang et al. / Int. J. Miner. Process. 74S (2004) S65–S74

Stehr, N., Herbst, J.A., 1984. Energy requirements and grinding Wang, Y., Forssberg, E., 1997. Ultrafine grinding and classification
kinetics for coal–water slurry preparation in high-speed stirred of minerals. In: Kawatra, S.K., Colorado, S.M.E. (Eds.),
ball mills. Department of metallurgy and metallurgical engineer- Comminution Practices, Chapter 27, ISBN: 0-87335-149-5,
ing, University of Utah, USA. pp. 203 – 214.
Stehr, N., Schwedes, J., 1983. Investigation of the grinding
behaviour of a stirred ball mill. Ger. Chem. Eng. 6, 337 – 343.