You are on page 1of 9


En route to multi-model scheme for clinker

comminution with chemical grinding aids
R. K. Mishra1, D. Geissbuhler2, H. A. Carmona1,3, F. K. Wittel1, M. L. Sawley2,
M. Weibel4, E. Gallucci4, H. J. Herrmann1, H. Heinz5 and R. J. Flatt*1
We present a multimodel simulation approach, targeted at understanding the behaviour of
comminution and the effect of grinding aids in industrial cement mills. On the atomistic scale, we
use molecular dynamics (MD) simulations with validated force field models to quantify elastic and
structural properties, cleavage energies as well as the organic interactions with mineral surfaces.
Simulations based on the discrete element method (DEM) are used to integrate the information
gained from MD simulations into the clinker particle behaviour at larger scales. Computed impact
energy distributions from DEM mill simulations can serve as a link between large scale industrial
and laboratory sized mills. They also provide the required input for particle impact fragmentation
models. Such a multiscale, multimodel methodology paves the way for a structured approach to
the design of chemical additives aimed at improving mill performance.
Keywords: Clinker, Comminution, Multimodel simulations, Molecular dynamics, Discrete element method, Grinding aid, Fragmentation, Ball mill

Introduction such that the size reduction is related to the energy con-
sumption.5,6 Such approaches are used for the design of
Comminution is a very energy intensive production step mills; however, parameters are only interpolated within
in the cement manufacturing, which involves the oper- the range of experience.7 What is missing is the physics
ation of large mills to achieve a desired clinker particle and chemistry behind the milling process. The problem
size distribution (PSD). In order to optimise this process, becomes particularly challenging by the interaction
we focus on physical and chemical aspects of grinding. between processes across length scales of 10 orders of
Industrial cement ball mills consist of a rotating dual magnitude difference, from the atomistic scale, where
chamber cylinder sized up to 8 m in diameter and 20 m surface formation and interactions rule, to the macro-
in length. Such mills are driven by electric motors in the scopic scale of the machinery involving the grinding media
range of 2040 MW, consuming 30% of the overall kinematics. The mill capacity is a function of the mill, the
electrical energy dedicated to the entire production chain liner design, the use of different grinding media,
of cement.14 Most of the earlier work on modelling of the charge and material grindability as well as of organic
clinker grinding was focused on the simulation of the grinding aids (GAs). Academic and industrial researchers
charge motion using the discrete element method alike have been faced with the problem of isolating
(DEM). Until now, numerical models based on different phenomena occurring on a reduced scale that are both
physical scales present in clinker and milling systems accessible to available analysis techniques and repre-
combining physicochemical aspects have not been sentative for the complete problem.
possible to integrate into a single methodology. Various In principle, the problem is too difficult to solve from
kinds of simulated systems described in this study con- a theoretical point of view in a closed form. Never-
tribute to construct a framework that captures clinker theless, we exemplify atomistic force field and (micro)-
grinding in an unconventional way. mechanical models based on the DEM on different
Comminution theory, as developed by process scales and explain how they can be enriched by findings
engineers, treats mills as black boxes, where input par- from models of smaller scales (see Fig. 1). In situ testing
ameters such as speed of rotation, loading, separator of GAs in industrial cement mills under stationary
efficiency and others are factors are combined in a way, conditions can only be reached after many tons of
clinker are milled, rendering screening studies both
costly and impractical. As a consequence, such exper-
Institute for Building Materials, ETH Zurich, Zurich 8093, Switzerland imental studies are often conducted in laboratory size
ACCES, School of Engineering, EPFL, Station 11, Lausanne 1015,
Switzerland tumbling mills (e.g. ball or planetary mills) containing
Departamento de Fisica, Universidade Federal do Ceara, Fortaleza grinding media that differ in size, shape and material.
60451-970, Brazil
Sika Technology AG, Zurich 8048, Switzerland
Since the ball kinematics may therefore not be com-
Department of Polymer Engineering, University of Akron, Akron, OH parable, it is unclear whether results obtained on these
44325 USA order of magnitude smaller mills still hold for the full
*Corresponding author, email scale cement mill. Obtaining suitable correspondence

2015 Institute of Materials, Minerals and Mining

Published by Maney on behalf of the Institute
Received 24 February 2015; accepted 30 April 2015
DOI 10.1179/1743676115Y.0000000023 Advances in Applied Ceramics 2015 VOL 114 NO 7 393
Mishra et al. Multi- -model scheme for clinker comminution with chemical grinding
g aids

1 Scales, objectives and methods involved in the multimodel scheme

is far from trivial, as high efficiency separators and hydrophobic ratios, molecular size etc.1,10 In this paper,
different process temperatures change the problem sig- the main implications of molecular dynamic (MD)
nificantly. Molecular modelling can give new insights, studies is to understand the interaction mechanisms of
which are not available experimentally or theoretically organic molecules with cement minerals. We are
but essential to understand working mechanisms of GAs involved in the development and validation of interface
for improved efficiency. force field parameters for both hydrated and unhydrated
The paper is organised as follows. First, we describe cement minerals that accurately reproduce density, lat-
simulations with GAs and clinker components in ato- tice constants, cleavage energy, solid/liquid interfacial
mistic force field m odels. S ubsequently, u pscaling of tension, as well as mechanical and structural properties
results over six orders of magnitude using fracture with respect to experiment.1012 Note that, compared to
mechanics8,9 leads to an idealised clinker model for alternative models,1315 our approach captures the
fragmentation simulations. This microscale model in- important surface polarity and interfacial properties
corporates the clinker as a multicrystal material and can correctly.
be used to simulate the outcome of typical situations for We focus on the four important phases of ordinary
which fragmentation in mills occurs. The computed Portland cement clinker (see Fig. 9), namely, tricalcium
kinematics of a (mesoscale) laboratory mill and full scale silicate (C3S) with Mg, Al and Fe impurities in the low
industrial cement ball mill, expressed in terms of impact per cent range (50270%), dicalcium silicate (C2S)
energy distribution (IED) of the particle charge, are then containing similar metal substitutions (15230%),
presented. Details are provided for further information tricalcium alumi-nate (C3A) in weight percentage of
required to be exchanged between different levels that 5210 and tetra-calcium aluminoferrite (C4AF) in weight
should provide a suitable basis for a complete im- percentage of 10215.16 Until now, we have validated
plementation of this multi-level, multi-model force field models of pure C3S (main phase of cement)
approach. Finally, we summarise and give an outlook and C3A (most reactive phase).10,12 We already
on possible future achievements in this field. concluded that surface polarity of C2S is lower than C3S
due to the absence of oxide ions and also due to more
Atomistic force field simulations of covalent character. We also found in a preliminary
simulation study that the cleavage energy of C2S is lower
clinker with GAs than that of C3S phase.17 Ferrite (C4AF) is one of the
Design of efficient GAs includes many factors such as most complex clinker phases in the cement and focus of
conformations of molecules on clinker surfaces, combi- cur-rent MD studies. Computed properties of C3S and
nation of polar functional groups, hydrophilic/ C3A phases are summarised in Table 1. The simulation

Table 1 Computed mechanical and surface properties of C3S and C3A phases of ordinary Portland cement clinker at
temperature 298 K10,12

Mineral Bulk modulus K/GPa Youngs modulus E/GPa Poissons ratio

C3S 1055 1525 (Ex), 1763 (Ey), 10311 (Ez) 0.280.08

C3A 983 1345 0.290.03
Cleavage energy/mJ m22 Temperature/K
C3S 134050 298
C3A 126050 298
Agglomeration energy/mJ m22 Typical grinding temperature/K
Hydroxylated C3S 24012 363
Hydroxylated C3A 25030 363

394 Advances in Applied Ceramics 2015 VOL 114 NO 7

Mishra et al. Multi- -model scheme for clinker comminution with chemical g
g rinding aids

2 Chemical structure of commercial GAs used in cement industry such as triisopropanolamine (TIPA), triethanolamine (TEA),
N-methyl-diisopropanolamine (MDIPA) and glycerine

3 Chemical structure of organic compounds, which are also tested for their effectiveness during clinker grinding such
as dipropylene glycol (DPG), dipropylene glycol monomethyl ether (DPGMME) and dipropylene glycol dimethyl
ether (DPGDME)

methodology and force field validations are discussed in and HO???O bond angles close to 180u. Additionally,
detail in our earlier papers on C3S and C3A.10,12 a weak hydrogen bond with a longer distance
Chemical GAs significantly improve mill efficiency, (0.25 nm+ 0.05) length and bond angle far from 180u is
characterised in terms of the specific surface area using found. The strength of calcium ion complexation with
Blaine fineness and sieve residues.1 They primarily shift alcohol based GAs depends upon the coordination number
the PSD curve towards smaller diameters but do not seem or number of nearest calcium ions. The adsorption strength
to influence the shape of the curve considerably. Because (in kJ mol21 molecule21) was found in the order TEA
of optical and practical restrictions, neither measure- (75+ 8) v TIPA(2100+ 15) * DPGDMEvMDIPA
ments in air nor in solvents delivers distribution curves (2121+ 18) * DPGMMEvDPG (151+ 12) * glycerine
beyond the scattering of laser PSD analysis. Using MD (2154+ 12) at 383 K temperature. Note that the values of
simulations, we relate adsorption and agglomeration the adsorption energy are not a measure of the performance
energies to rank the GAs on the basis of interaction of a GA (the correct measure is agglomeration energy) and
strengths with clinker surfaces as well as reduction in also do not correlate with the volatility of the organic
agglomeration energies. The agglomeration energy is the compounds (a possible measure is the cohesive energy of the
released energy when freshly cleaved surfaces (particles) pure liquid).
come together in order to reduce specific surface area and We computed agglomeration energies of dry and
thus cleavage/surface energy.10 Grinding aids (mostly hydroxylated clinker surface (C3S, C3A) without and with
organic compounds) reduce the surface energy of the GAs. Figure 4b and c gives numbers of agglomeration
clinker and therefore also the energy that is released when energies of test and commercial GAs with chemical
particles agglomerate. The energy consumption decreases dosages of 0.20 mg m22. Experimental findings also
during grinding as a consequence of various physical and follow trends on agglomeration energy.1,10,12 We pre-
chemical phenomena happening simultaneously during sented the experimental comparison of commercial GAs
the clinker milling in the presence of GAs.1 on the basis of normalised Blaine fineness and sieve resi-
The chemical structure of selected commercial GAs due at 32 mm with respect to data of MDIPA in Table 2.
[triisopropanolamine (TIPA), triethanolamine (TEA), Correlation of agglomeration energy with Blaine fineness
N-methyl-diisopropanolamine (MDIPA) and glycerine] and sieve residue of clinker particles is quite consistent to
and other organic compounds [dipropylene glycol (DPG), predict the performance of GAs. Grinding aids help to
dipropylene glycol monomethyl ether (DPGMME) and reduce surface forces between cleaved clinker particles,
dipropylene glycol dimethyl ether (DPGDME)] is shown in which are very well captured by agglomeration studies.
Figs. 2 and 3. Models of hydroxylated C3S and C3A surfaces These atomistic studies give an important insight into
are used for the simulations because the majority of com- the molecular interactions; however, the direct upscaling of
mercial clinkers are mostly hydroxylated.1,10,12 Interactions results is only partially possible. For sure, the mechanical,
of DPG on hydroxylated C3S surface is exemplarily shown structural and surface properties of clinker phases can be
in Fig. 4a. The hydroxyl groups of the hydroxylated C3S embedded in an artificial clinker model that is used in DEM
surfaces are close to the polar groups of organic molecules. simulations for clinker particle fragmentation. For the
Adsorption of GAs occurs on the ionic clinker surfaces ball mill simulations with clinker particles, computed
mostly due to hydrogen bonds, complexation of superficial agglomeration energies can be important information.
calcium ions and dipolar interactions. Dipropylene glycol
monomethyl ether, DPG, TEA, TIPA, MDIPA and
glycerine are adsorbed more due to hydrogen bond for- Fragmentation of artificial clinker
mation. For these molecules, we find strong hydrogen We simulated composite breakable particles with crys-
bonds due to shorter O???H bond length (0.17 nm+ 0.03) talline and amorphous phases with respective material

Advances in Applied Ceramics 2015 VOL 114 NO 7 395

Mishra et al. Multi- -model scheme for clinker comminution with chemical g
g rinding aids

4 a dipropylene glycol molecule on hydroxylated C3S surface with hydrogen bonds forming between hydroxyl groups of
molecules to silicate and hydroxide groups on C3S surface at ball mill temperature 383 K; b computed agglomeration energy
of dry C3S, hydrated C3S and in presence of DPG, DPGMME and DPGDME molecules on hydrated C3S surfaces at typical
grinding temperature 363 K; and c reduced agglomeration energy due to presence of commercial GAs with chemical
dosage of 20 mg m22 for C3A and C3S;12 effect of additives is similar with trend towards slightly higher agglomeration energy
for C3S
Adsorption energy: HCTEA,HCTIPA,HCMDIPA,HCglycerine; agglomeration energy: C3S.HC.HCglycer-
ine.HCTEA.HCTIPA.HCMDIPA; grinding performance: clinker,HC,HCglycerine,HCTEA,HCTIPA,HCMDIPA

properties (Table 3) using the DEM. Particle models with intrinsic cleavage planes. As multiple crystallites are
cohesive beam truss elements are ideal for studying involved, first, a packing of evolved single crystallites is
dynamic fracture and fragmentation problems. constructed (Fig. 5a) from a collection of single crystals
As various processes take place instantaneously during following the grain size distribution of C3S from a repre-
a short time span, the collaborative dynamics of particles, sentative clinker microscopic section. Interstitial spaces
considering Hertzian collisional repulsive contacts with are then filled by randomised particles to resemble an
friction, cohesive interaction and volumetric forces, is amorphous phase.22 Finally, a spherical sample is cut
solved by forward integration of the equation of motion of from the system and impacted at different energy levels
particles. Various fragmentation mechanisms, their against a rigid target. In calibration simulations, we match
origin, evolution and interaction during the process were macroscopic elastic properties and fracture energies to the
studied in detail.1821 Clearly, clinker is a multiphase and values obtained from force field calculations.
multicrystalline material, calling for a consideration of the Beams are removed once their elliptical breaking rule
heterogeneity of some sort. We simplify it by considering a of the von Mises criterion type
single crystalline phase for C3S and C2S with elements
organised in a hexagonal close packing, exhibiting 1=1th 2 max jhi j; jhj j=hth $ 1 1

Table 2 Normalised Blaine fineness and sieve residue at 32 mm with respect to MDIPA values; data were generated with help
of grinding experiments in a laboratory mill

Dosage of Concentration of Relative Blaine Relative sieve

Compound GA/g ton21 cement GA in water/wt-% fineness/wt-% residue at 32 mm/wt-%

MDIPA 500 40 100.00 100.00

TIPA 500 40 95.80 107.80
TEA 500 40 92.70 112.10
Glycerine 500 40 90.50 118.50

396 Advances in Applied Ceramics 2015 VOL 114 NO 7

Mishra et al. Multi- -model scheme for clinker comminution with chemical g
g rinding aids

Table 3 Discrete element method model properties for fragmentation of artificial clinker

Microscopic property Symbol C3S C3A Unit

Stiffness Eb 296 202 GPa
Average length Lo 5 5 mm
Diameter Db 3 3 mm
Strain threshold 1th 0.025 0.02 ...
Bending threshold hth 3.0 2.5 8
Shape parameter k 10 10 ...
Stiffness Ep 296 202 GPa
Diameter Dp 5 5 mm
Density r 5000 5000 kg m23
Friction coefficient m 1 ...
Damping coefficient cn 1027 Ns m21
Friction coefficient ct 1028 NS m21
Time increment Dt 1610210 s
Number of elements Np 90 000 ...
Number of beams Nb 700 000 ...
System diameter D 260 (spherical samples) mm
Macroscopic property
Stiffness E 14510 13510 GPa
Poissons ratio n 0.1 0.14 ...
Density r 2700 3500 kg m23
Volume fraction 78 22 %
Ultimate strength smax 2.4 1.7 GPa

is fulfilled, where 1 is the longitudinal strain, and hi and robust with respect to the internal structure of the
hj are the largest general end beam rotation angles material. Impact simulations were conducted for differ-
connecting particle i with j. Thresholds values 1th are ent collision energies. For the particular size of samples
sampled from a Weibull distribution. The macroscopic we simulated, we find that a critical collision energy of
strength of the material can be tuned by adopting the 0.8 mJ is necessary to fragment the system. Further
average breaking thresholds in each material phase increase in the impact energy produces more fragments
separately so that the ratio between the ultimate strength with smaller sizes, but usually leaves at least one large
of C3S and C3A follows one of the respective surface fragment. For all configurations, a sharp transition from
energies determined by MD simulations. Additionally, the damage to the fragmentation regime is observed,
the breaking thresholds were set so that the ultimate with critical collision energies being smallest for mono-
strength of a microsample C3A with dimensions in the crystalline samples. The mass distribution of the frag-
range of the characteristic grain size is near 2 GPa. ments follows a power law Pam 2a with an exponent
Damage initiates inside spherical samples above the a < 1:6 for small fragments that is characteristic for
contact zone in a region where the circumferential stress the branchingmerging process of unstable crack
field is tensile. Cracks initiated in this region grow to growth. Moreover, this exponent depends only on the
form meridional planes. If the collision energy exceeds a dimensionality of the system and not on the micro-
critical value that depends on the materials internal structure (see Fig. 5c). Hence, attempts to reduce energy
structure, cracks reach the sample surface resulting in in comminution by manipulating the clinker micro-
fragmentation (see Figs. 5b and 6). We discovered that structure have a rather limited potential with respect to
this primary fragmentation mechanism is surprisingly this fragmentation case.

5 a crystalline phase for alite (C3S) and amorphous phase for aluminates (C3A) and others, b two largest fragments for
multiphase samples (red bonds resemble amorphous crystalline, green ones intercrystalline interfaces) and c fragment
mass distribution for different microstructures18,22

Advances in Applied Ceramics 2015 VOL 114 NO 7 397

Mishra et al. Multi- -model scheme for clinker comminution with chemical g
g rinding aids

6 Top view of slices of sample at different y-positions, from y 5 85mm (top left) to y 5 250mm (bottom right); bond colours
correspond to material properties: black and red, C3A; blue and green, C3S; white are broken bonds

Energy dependent PSDs for impact fragmentation are comminution process.28,29 In the proposed multiscale
important for the simulation of the outcome of ball numerical approach, laboratory mills are best used to
mills; however, one should note that fragmentation by study the evolution of the PSD based on the previously
impact above the critical energy is nothing but one described microscale breakage models. In the present
possible way for size reduction of clinker as simulations study, particle breakage has not been modelled in the
of IEDs reveal. Attrition, impact fatigue and fragmen- mill simulations. Of particular importance in this study,
tation inside the granular bed with multiple contact however, is the determination of the IED, which not
points resemble cases that need to be studied separately only provides a basis for comparison between mills of
to provide further realistic fracture kernels. In addition, different sizes (i.e., the scaling of results)30,31 but also a
it needs to be noted that larger microsections of clinker means to link to the microscopic modelling of the
exhibit voids and cracks that point at large residual breakage of individual clinker particles.
stresses that were not considered at this time.
Planetary mill
Discrete element method simulations of the particle
Simulation of ball mills dynamics in both planetary and laboratory scale ball
Simulations with DEMs have been employed by mills have been carried out. A spring dashpot model32
numerous authors to study particle dynamics in a variety has been employed to determine the normal and tan-
of tumbling mills.2327 While a cement mill operator is gential collisional forces and torques acting on the par-
principally interested in global behaviour (e.g. particle ticles. Presented here are results for a commercial
size evolution during mill operation), this is determined planetary mill, the Fritsch Pulverisette 6, which is a
by phenomena occurring at the microscopic scale, as small but convenient device for studying comminution.
described in earlier sections. To gain insights into how This device has a grinding bowl of 100 mm diameter
the presence of GAs influences the efficiency of the rotating at 220 rev min21 and mounted on a disc coun-
comminution process in an industrial cement mill, both ter-rotating at 400 rev min21.23 The ball charge consists
a full scale ball mill and laboratory mills have been of 25 steel balls of 20 mm diameter. The operation of a
considered. Although we are ultimately interested in planetary mill can be modified by changing the ro-
predicting the behaviour of a full scale mill, laboratory tational speed (while maintaining the same ratio of bowl
mills are considerably more flexible and accessible for and disc speeds), by changing the size and number of
both experimental and numerical studies. Such mills are steel balls and by modifying the clinker charge.
therefore used extensively to obtain detailed exper- The purpose of DEM simulations was to analyse the
imental measurements that can provide insights into the effect of such operational changes. Figure 7a shows an

398 Advances in Applied Ceramics 2015 VOL 114 NO 7

Mishra et al. Multi- -model scheme for clinker comminution with chemical g
g rinding aids

example of the instantaneous charge position and IEDs and including only the ball charge (i.e. neglecting clinker
computed for a planetary mill containing 100 g of clinker particles).25,26 These reduced DEM simulations can,
particles. The corresponding computed energy spectra are nevertheless, provide representative IEDs for a large scale
presented in Fig. 7b, where the dissipated energy is binned cement mill. Figure 8 presents a comparison of the IEDs
logarithmically with 20 bins/decade. Low energy collisions computed for the two chambers of the cement plant mill
between clinker particles and high energy collisions invol- and for a planetary laboratory mill. It can be noticed that
ving ball clinker contacts can be associated with the the first chamber of the cement mill provides the higher
observed two peaks. This information can provide the collision energies necessary to break large clinker pieces,
impact energy required as input to the particle breakage while the second chamber enables finer grinding at lower
modelling described in fragmentation study, enabling a collision energies. It can also be observed from Fig. 8 that,
prediction of the evolution of the PSD due to comminu- by varying the rotational speed of the planetary mill, the
tion. Preliminary studies for the planetary mill indicated peak of the collisional energy spectrum can be tailored to
this approach to be promising, although its accuracy match that of the cement mill. Since the energy dissipated
remains to be confirmed by validation studies. in clinker particle collisions is central to particle size re-
duction, it is reasonable to conjecture that comminution
Cement mill characteristics of the planetary mill will then be repre-
An industrial cement mill generally consists of two sentative of those of the full scale cement mill. The
chambers, the first containing large balls and a lifting liner, resulting evolution of the PSD can then be predicted based
and the second containing smaller balls and a classifying on the results of breakage simulations proposed in particle
liner. For the present study, a mill of 4.4 m diameter fragmentation.
rotating at 15.2 rev min21 (*75% of critical speed) was
considered. Although DEM simulations of the complete
mill operation are not feasible with current computational
Summary and outlook
resources, these requirements can be reduced by con- We have combined diverse multiscale modelling approa-
sidering a short axial slice of each of the mills chambers ches to obtain new insights into the complex industrial

7 a ball charge, with clinker particles coloured by their speed (blue: 0; red: 3 m s21); b impact energy distributions for normal,
shear and total collisions

8 Impact energy distributions computed for two chambers of cement plant ball mill and for planetary mill for three different
disc rotational speeds

Advances in Applied Ceramics 2015 VOL 114 NO 7 399

Mishra et al. Multi- -model scheme for clinker comminution with chemical g
g rinding aids

9 Crystal structure of clinker phases Ca M3 polymorph of tricalcium silicate (16261 unit cells, a 1.2235 nm); 33
b monoclinic dicalcium silicate of stable b polymorph (26261 unit cells, a 0.5502 nm);34 c unit cell of cubic lattice
lattice of tricalcium aluminate (16161 unit cell, a 1.5263 nm);35 d super cell of tetracalcium aluminoferrite, super cell of
tetracalcium aluminoferrite, (26162 unit cells, a 0.53369 nm) 36

problem of clinker comminution. Atomistic MD simu- problem and studying them individually.
lations quantitatively show that reducing the agglomera- As demonstrated in the present paper, atomistic simu-
tion energy is the main function of GAs. Microscale DEM lations can not only play an important role in deter-
simulations of particle fragmentation with multiphase mining the relative efficiency of different GAs but can
properties obtained from the atomistic simulations give also give quantitative information on properties of
fragmentation thresholds and insight into the role of clinker phases and their surface interactions. For an
different phases. For impact fragmentation, crack planes ideal stress free clinker, this would suffice. However, in
are not substantially deflected at grain or phase bound- future works, disorder and quenching of clinker should
aries, which is important with the reactivity of the cement be simulated to relax the assumption of stress and defect
in mind. Laboratory mills (especially the planetary mill) free systems. Additionally, studies of attrition, fatigue
provide a convenient framework to couple the breakage and bed fragmentation would enrich the fragmentation
dynamics of individual clinker particles with the global scenarios available for the more detailed simulation of
IED computed using mesoscale DEM simulations. On the comminution in cement mills.
largest scale, the behaviour of representative axial slices of
the two chambers of a cement mill has been computed,
proving that impact spectra can be comparable with those
from laboratory mills. To summarise, the energy spectra We acknowledge support by the Swiss Commission for
obtained from DEM simulations provide a quantitative Technological Innovation under grant no. KTI 13703.1
measure that forms the basis of scaling the behaviour of a PFFLR-IW and by the Petroleum Research Fund of the
laboratory mill to that of an industrial mill. American Chemical Society (grant no. 54135-ND10).
A holistic numerical simulation methodology for
industrial cement mills, including grinding media and
fragmenting charge, will be most likely unattainable for References
a number of years. Nevertheless, todays computers are
sufficiently powerful to identify and analyse important 1. M. Weibel and R. K. Mishra: Comprehensive understanding of
grinding aids, ZKG Int., 2014, 67, (6), 2839.
physical phenomena occurring at very different length 2. N. Madlool, R. Saidur, M. Hossain and N. Rahim: A critical review
scales. These phenomena can be studied in more detail on energy use and savings in the cement industries, Renewable and
by isolating the essential aspects from the global Sustainable Energy Rev., 2011, 15, (4), 20422060.

400 Advances in Applied Ceramics 2015 VOL 114 NO 7

Mishra et al. Multi- -model scheme for clinker comminution with chemical g
g rinding aids

3. A. Atmaca and M. Kanoglu: Reducing energy consumption of a 20. F. Kun and H. J. Herrmann: Transition from damage to
raw mill in cement industry, Energy, 2012, 42, (1), 261269. fragmentation in collision of solids, Phys. Rev. E, 1999, 59E, (3),
4. D. Fuerstenau and A.-Z. Abouzeid: The energy efficiency of ball 26232632.
milling in comminution, Int. J. Miner. Process., 2002, 67, (1), 161185. 21. F. K. Wittel: Single particle fragmentation in ultrasound assisted
5. H. Masuda, K. Higashitani and H. Yoshida: Powder technology: impact comminution, Granular Matter, 2010, 12, (4), 447455.
fundamentals of particles, powder beds, and particle generation; 22. I. Nikolakopoulos: Clinker simulation, MS thesis, ETH Zurich,
2006, Boca Raton, FL, CRC Press. Zurich, Switzerland, 2013.
6. A. D. Salman, M. Ghadiri and M. Houndlow: Handbook of 23. D. Geissbuhler and M. L. Sawley: Particle motion and energy
powder technology, Vol. 12, Particle breakage; 2007, dissipation spectra in a planetary ball mill. Proc. 3rd Int. Conf.
Amsterdam, Elsevier Publishing. on Particle-based methods fundamentals and applications,
7. Y. Kanda and N. Kotake: Comminution energy and evaluation in eds. M. Bischoff et al.,,, Stuttgart, Germany, September 2013,
fine grinding, in Handbook of powder technology, 236246.
(ed. A. D. Salman et al.), Vol. 12, 529550; 2007, Amsterdam, Elsevier. 24. B. Mishra: A review of computer simulation of tumbling mills by
8. R. Phillips: Multiscale modeling in the mechanics of materials, the discrete element method, Part II - practical applications, Int.
Cur. Opin. Solid State Mater. Sci., 1998, 3, (6), 526532. J. Miner. Process., 2003, 71, (1), 95112.
9. D. Holland and M. Marder: Ideal brittle fracture of silicon studied 25. P. W. Cleary: Ball motion, axial segregation and power
with molecular dynamics, Phys. Rev. Lett., 1998, 80, (4), 746. consumption in a full scale two chamber cement mill, Miner.
10. R. K. Mishra, R. J. Flatt and H. Heinz: Force field for tricalcium Eng., 2009, 22, (9), 809820.
silicate and insight into nanoscale properties: cleavage, initial 26. P. W. Cleary and R. D. Morrison: Understanding fine ore
hydration, and adsorption of organic molecules, J. Phys. Chem. breakage in a laboratory scale ball mill using DEM, Miner.
C, 2013, 117C, (20), 1041710432. Eng., 2011, 24, (3), 352366.
11. H. Heinz, T.-J. Lin, R. K. Mishra and F. S. Emami: 27. N. Weerasekara, M. Powell, P. Cleary, L. M. Tavares,
Thermodynamically consistent force fields for the assembly of M. Evertsson, R. Morrison, J. Quist and R. Carvalho: The
inorganic, organic, and biological nanostructures: the INTERFACE contribution of DEM to the science of comminution, Powder
force field, Langmuir, 2013, 29, (6), 17541765. Technol., 2013, 248, 324.
12. R. K. Mishra, L. Fernandez-Carrasco, R. J. Flatt and H. Heinz: 28. C. F. Burmeister and A. Kwade: Process engineering with
A force field for tricalcium aluminate to characterize surface planetary ball mills, Chem. Soc. Rev., 2013, 42, (18), 76607667.
properties, initial hydration, and organically modified interfaces in 29. S. Rosenkranz, S. Breitung-Faes and A. Kwade: Experimental
atomic resolution, Dalton Trans., 2014, 43, (27), 1060210616. investigations and modelling of the ball motion in planetary ball
13. R. T. Cygan, J.-J. Liang and A. G. Kalinichev: Molecular models of mills, Powder Technol., 2011, 212, (1), 224230.
hydroxide, oxyhydroxide, and clay phases and the development of a 30. J. A. Herbst: A microscale look at tumbling mill scale-up using high
general force field, J. Phys. Chem. B, 2004, 108B, (4), 12551266. fidelity simulation, Int. J. Miner. Process., 2004, 74, S299S306.
14. A. G. Kalinichev and R. J. Kirkpatrick: Molecular dynamics 31. T. Iwasaki, T. Yabuuchi, H. Nakagawa and S. Watano: Scale-up
modeling of chloride binding to the surfaces of calcium methodology for tumbling ball mill based on impact energy of
hydroxide, hydrated calcium aluminate, and calcium silicate grinding balls using discrete element analysis, Adv. Powder
phases, Chem. Mater., 2002, 14, (8), 35393549. Technol., 2010, 21, (6), 623629.
15. R. Shahsavari, R. J. M. Pellenq and F.-J. Ulm: Empirical force 32. P. A. Cundall and O. D. Strack: A discrete numerical model for
fields for complex hydrated calcio-silicate layered materials, granular assemblies, Geotechnique, 1979, 29, (1), 4765.
Phys. Chem. Chem. Phys., 2011, 13, (3), 10021011. 33. W. Mumme: Crystal structure of tricalcium silicate from a
16. H. F. W. Taylor: Cement chemistry; 1997, London, Academic Press. Portland cement clinker and its application to quantitative
17. R. K. Mishra, R. J. Flatt and H. Heinz: A force field for XRD analysis, Neues Jahrb. Mineral. Monatsh., 1995, 4, 145160.
dicalcium silicate and evaluation of surface properties, 34. K. Jost, B. Ziemer and R. Seydel: Redetermination of the structure of-
In press. dicalcium silicate, Acta Crystallogr. B, 1977, 33B, (6), 16961700.
18. H. A. Carmona, A. V. Guimaraes, J. S. Andrade, I. Nikolakopoulos, 35. P. Mondal and J. Jeffery: The crystal structure of tricalcium
F. K. Wittel and H. J. Herrmann: Fragmentation processes in aluminate, Ca3Al2O6, Acta Crystallogr. B, 1975, 31B, (3), 689697.
two-phase materials, Phys. Rev. E, 2015, 91E, (1), 012402012407. 36. G. J. Redhammer, G. Tippelt, G. Roth and G. Amthauer:
19. H. A. Carmona, F. K. Wittel, F. Kun and H. J. Herrmann: Structural variations in the brownmillerite series Ca2(Fe2
Fragmentation processes in impact of spheres, Phys. Rev. E, xAlx)O5: single-crystal X-ray diffraction at 25 degrees
2008, 77E, (5), 051302. C and high-temperature X-ray powder diffraction (25 degrees
C,T,1000 degrees C), Am. Miner., 2004, 89, (23), 405420.

Advances in Applied Ceramics 2015 VOL 114 NO 7 401