Академический Документы
Профессиональный Документы
Культура Документы
Powder Technology
a r t i c l e i n f o a b s t r a c t
Article history: The paper initially focusses on developing a test procedure for measuring the wear of grinding balls and then
Received 21 June 2017 correlating the results with the wear obtained in industrial scale cement ball milling applications. The cost of
Received in revised form 31 October 2017 grinding media is one of the major items and it needs to be optimized by testing different types of media.
Accepted 11 November 2017
Since industrial scale trials are time consuming, there is a need for an easy, accurate and fast media wear testing
Available online 15 November 2017
procedure for cement ball milling applications. Such a procedure enables cement plants to compare wear results
Keywords:
of different types of media and choose the optimal one for the economy of the grinding operation. Within the
Media wear scope of the study, 6 types of 30 mm media having different chemical compositions were subjected to wear
Media consumption measurements at different time intervals by utilizing laboratory scale ball mills. The studies showed that repro-
Wear ducible results can be obtained from the test apparatus hence the methodology is said to be consistent. The wear
Cement results concluded that the order of the media wear from the highest to the lowest was the same independent of
Dry grinding the mill dimension and whether the material is inside the mill. The results also implied that the chemical
Optimization composition or the microstructure of the media was an inuencing parameter on the wear therefore there is
room for improvement with regards to the quality of media. This study proved that the media wear measured
at industrial scale cement ball mills was in good agreement with the laboratory scale results.
2017 Elsevier B.V. All rights reserved.
1. Introduction Literature reports that there are 3 basic wear mechanisms that are
abrasion, impact and corrosion, which can be correlated with the mass
In a comminution process, the choice of the grinding media is of losses of the grinding media. In a grinding system, owing to the nature
great importance. Therefore, wide range of products should be com- of the operation, the complex synergies of these mechanisms can arise
pared and the economy of the application i.e., cost of media, media [2]. For instance, decreasing rotational speed of the ball mill results in
wear, grinding performance of media etc. should be evaluated in details. cascade action or abrasion wear mechanism be more dominant hence
In the economy of the grinding operation, media consumption is one of the inuence of impact mechanism is reduced [4]. Consequently, the
the primary items. Literature reports that the wear of liners together operating conditions of the mill should be considered initially and
with the grinding media accounts for about 1525% of the mining oper- then the focus should be given on media selection.
ating costs [1]. The wear costs of the grinding media given in the litera- Until so far, the variations in media wear have been attributed to
ture is mostly collected from wet grinding operations. There is limited several parameters. In wet milling, pulp coating on ball surfaces,
data on dry cement grinding circuits. However, during communications which is the function of rheology, reduces the media wear [2]. It can
with the plant people it is found out that grinding media consumption be concluded that the solid loading and viscosity modiers can
constitutes 38% of the total operating comminution costs depending inuence the wear characteristics of the media [5]. During the wet
on the composition of the media and the type of cement production. milling, the resistance of media to corrosive action should also be
Consequently, it is important to make thorough and reliable evaluations considered. The wear of the media is supposed to be increasing as a
on media consumption so that the overall operating costs can be opti- result of the interactions between corrosive and abrasive components
mized [2]. In this context, long term observations should be made and [6]. Hardness of the media is another parameter. In case of having
detailed studies should be conducted for specic applications so that increased hardness number, the media resists more to impact mecha-
the behaviour of the media can be understood more clearly [3]. nism rather than the abrasion [2]. As a conclusion of the media wear
studies, although the individual effects of these parameters have
Corresponding author. been summarized, which of the action is more predominant still
E-mail address: okyaltun@hacettepe.edu.tr (O. Altun). remains controversial [7].
https://doi.org/10.1016/j.powtec.2017.11.033
0032-5910/ 2017 Elsevier B.V. All rights reserved.
S.S. olak et al. / Powder Technology 325 (2018) 678686 679
Table 1
The chemical compositions of the media tested.
also help predicting the time necessary to recharge the media into the A B C D 0 1 2 3
mill. Such a procedure can improve the control on the milling operation A1 61 62 61 61 57 58 59 61
to ensure the sustainability of the overall process. A2 61 61 60 61 58 58 59 60
A3 61 62 60 61 57 58 59 60
B1 60 61 61 61 59 59 60 61
2. Materials and methods
B2 60 61 61 61 58 59 60 60
C1 61 62 61 61 57 57 58 60
2.1. Media types
Fe 0.93
SiO2 33.98
Table 2 Al2O3 13.03
The volumetric fractions (%) of the elements found in the microstructure. CaO 29.87
MnO 2.37
Name Martensitic matrix Primary carbides Secondary carbides
MgO 12.18
A1 60 40 6 P2O5 b0.01
A2 50 50 5 SiO2 0.65
A3 60 40 15 Na2O 0.37
B1 40 60 6 K2O 0.84
B2 40 60 12 TiO2 0.9
C1 50 50 10 Bond work index (kWh/t) 20.01
680 S.S. olak et al. / Powder Technology 325 (2018) 678686
Table 5 Table 7
The chemical composition and Bond work index value of the The dimensions of the mills utilized.
clinker.
Mill 1 Mill 2 Mill 3
Loss on ignition (%) 0.24
Mill diameter (mm) 305 711 950
Insoluble residue (%) 0.08 Mill length (mm) 305 508 1450
CaO (%) 66.05 % Critical speed 87 60 27
SiO2 (%) 21.59
Al2O3 (%) 6.16
Fe2O3 (%) 2.67
MgO (%) 1.48 2.3. Test apparatus
SO3 (%) 0.42
Na2O (%) 0.41 To investigate the effect of the mill geometry on the wear rates,
K2O (%) 0.98 tumbling mill tests were carried out by utilizing the mills having 3
Cl (%) 0.01
FCaO (%) 1.11
different geometries (Table 7). In these tests, the surface area and
Bond work index (kWh/t) 13.99 weight of the ball charge were kept constant as the controlled variables,
and the relationship between the mill geometry and the wear rates was
investigated. During the test work, each of the mill was charged with
10 kg of media and subjected to tumbling action at different time
understood from the volume hardness values that the hardness varies intervals i.e., 60, 180, 420, 900 min.
with distance such that the core of the media (point 0) has the lowest After completing each of the tumbling periods, the whole of the ball
hardness value and it increases steadily through the surface. charge was taken out of the mill, well-cleaned individually and then
weighed. In the end of the study, the cumulative weight loss of the
media as a function of time was plotted as illustrated in Fig. 3. Such a
2.2. Material characterization plot enabled comparing different media types hence the inuences of
the media properties or the milling conditions were investigated.
The media wear results were obtained for with and without material Industry and literature expresses the wear rate of media as;
cases. The tests without the material focussed only on the inuences of
the media characteristics on the wear. On the other hand, the tests with Consumption by unit of energy consumed, (kg/kWh)
the material considered the effects of the material characteristics as Consumption by unit of ore ground, (kg/ton) [11]
well. Within the scope, the tests with the material were performed by
using the granulated blast furnace slag, which is known as highly Within the study, the results were expressed by both units by
abrasive, clinker and copper ore, obtained from the cement plants and beneting from the Bond's energy equation (Eq. (1)) [12],
copper producer. The chemical composition, Bond work indices which
were obtained by applying conventional Bond test procedure pre-
1 1
scribed by Bond [10] and particle size distributions of the materials W 10 Wi pp 1
used in the tumbling mill with material tests were summarized and P80 F80
illustrated in Tables 4, 5 and 6 and Fig. 2.
Eq. (1) requires the particle size distributions of the feed and product
as well as the Bond work indices of the material to calculate the power
draw of the mill. Consequently, for the test with the material, ground
Table 6 material was also taken out of the mill after the test and the size distri-
The chemical composition and Bond work index value of the cop- bution was determined then inputted into the equation to calculate the
per ore. work done by the laboratory mill. Afterwards the ratio of wear result to
Cu (%) 3.28 the work done by the mill was calculated.
Zn (%) 2.34
Pb (%) 0.17 2.4. Industrial media wear measurements
Bond work index (kWh/t) 12.5
The study was not limited only by the laboratory scale measure-
ments. Since the aim was to develop correlation with the industrial
100
0.07%
Cumulative Undersize %
80 0.06%
Cum. Mass Loss %
0.05%
60
0.04%
40 0.03%
0.02%
20 Clinker
Granulated Blast Furnace Slag 0.01%
Copper
0 0.00%
100 1000 0 5 10 15 20
Particle Size (m) Time (hours)
Fig. 2. The particle size distributions of the feed samples. Fig. 3. An example of the media wear plots obtained after the test work.
S.S. olak et al. / Powder Technology 325 (2018) 678686 681
Table 8
The technical specications of the ball mills and the production data of the grinding circuits.
scale results, in the nal stage of the study, long-term observations 3. Results and discussions
(6 months) were made at 3 different cement plants during CEM I
42.5R production where one of the tested grinding media was charged 3.1. Reproducibility of the test results for the experimental apparatus
(A1 type of media). The ow sheet of the cement grinding circuits and
the specications of the units are summarized in Fig. 4 and Tables 8 Prior to performing series of media wear test works, reproducibility
and 9. of the experimental apparatus was statistically tested with the aim of
As a result of the long-term observations, the average wear rates of improving the condence on the data obtained. At rst, it was proved
the media were determined by the plant people as 23, 50 and 13 g per that the data was normally distributed within the level of 95%
ton of cement produced for Plants A, B and C respectively. Variations condence by implying Shapiro-Wilk test [13]. Once the normality of
in the characteristics of the feed material (size, hardness, moisture the data was determined, t-test (paired two sample for mean) [14]
content, etc.) during the period in which the observations were made was applied and the null hypothesis was tested. The test works were
are thought to inuence the wear rates. performed at two different days at the same milling conditions and A1
Table 10
Table 9 The milling conditions of the reproducibility tests.
Average ball sizes of the ball mills.
Day 1 Day 2
Average ball size (mm)
Duration of test (mins) 150, 300, 450, 600
Cement plant A 36 Ball diameter (mm) 30
Cement plant B 39 Ball lling ratio (%) 10
Cement plant C 36 Material lling ratio (%) 10
682 S.S. olak et al. / Powder Technology 325 (2018) 678686
Table 11
The results of the reproducibility tests.
Table 12
The results of the statistical evaluations.
Day 1 Day 2
type of media was charged into the mill with granulated blast furnace
slag. After the test works both the wear rates of the media and particle
size distributions were measured. The test conditions are summarized
in Table 10.
As can be understood from Table 10 cumulative tumbling time
reached to 600 min. Following the test runs, mass loss of the media
was measured for each time interval and the data was compared
(Table 11).
The reproducibility of the test apparatus was assessed by consider-
ing the differences in the mean values of the two days (Table 12). In Fig. 7. Media wear results obtained from Mill 2.
this context, the hypothesis whether there was a signicant difference
between the values was tested. As a result, it was found that two-tail
probability value (P) was greater than t-statistic value hence the hy- was just tumbled in 3 different mill dimensions (Table 7). As a result,
pothesis was rejected meaning that there was no signicant difference these tumbling tests investigated only the inuence of the chemical
between the results. As a conclusion, the experimental apparatus is composition of the media as well as the geometry of the mill on the
said to be producing repeatable results at the same milling conditions. wear results. The test results are illustrated in Figs. 6, 7 and 8.
In addition to the media wear results, the particle size distributions Fig. 6 illustrates that the rate of media wear starts losing its linear
of the products were also compared. Fig. 5 implies that after 600 min trend after 3 h of tumbling time for A2, A3, B1, B2 and C1 types of
of grinding time, the products of Day 1 and Day 2 are similar to each media. However, for Mill 2 and Mill 3 different behaviour was observed
other. All these studies proved that both media wear results and size since the rate of media wear had exponential trends. This can be attrib-
distributions can be used for further evaluations. uted to the effect of the mill dimension, as the media in Mill 2 and Mill 3
had higher potential energy than that of Mill 1.
3.2. The results of the tumbling tests (without material) The results of the tumbling tests also implied that the wear rate of
media depended on the mill dimensions. As the mill dimension in-
In a grinding operation, it is known that the material has cushioning creased, higher rate of wear was obtained. Fig. 9 illustrates the variation
effect on the media and increasing the material load reduces the media of the media wear with the mill dimension after 15 h of tumbling time.
wear [6]. Within the study, the media wear tests were initially per- Another conclusion from the tumbling tests was even if the mill
formed for the case without a material in the mill. The 10 kg of media dimension changed, wear results from the highest to the lowest were
in the same order; A1 N A2 N B1 N A3 N C1 N B2. In other words, A1
type of media had the highest rate of wear where the B2 type of
Fig. 5. The particle size distributions of the product after 600 min of grinding time. Fig. 8. Media wear results obtained from Mill 3.
S.S. olak et al. / Powder Technology 325 (2018) 678686 683
Fig. 10. The variation of the cumulative mass loss (%) with the tumbling time.
Fig. 9. The variation of the media wear with the mill dimension.
Table 14
The calculation table for 10% material lling ratio.
Table 16
Test conditions of the media wear measurements with material.
higher Bond work index value. As a result of the material tests it can
be concluded that the developed methodology is sensitive to the varia-
tion in the material characteristics.
3.4. The relationship between the micro-structure of the media and the
Fig. 12. The variation of the cumulative mass loss (kg/kWh) with the tumbling time. wear results
0.040%
10%, 0.033%
0.035% 20%, 0.029%
Average Mass Loss (%)
0.025%
y = -0.0003x + 0.0004 50%, 0.021%
0.020% R = 0.9677
Fig. 14. The variation of the wear rate with the type of media (with material).
0.015% 60%, 0.017%
0.010%
80%, 0.007% Table 17
The milling conditions of the material tests.
0.005%
100%, 0.004%
Type of mill Mill 1
0.000%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Type of material Clinker, copper ore and slag
Material Filling Ratio Type of media A1
Material load (%) 80
Media load (kg) 10
Fig. 13. The variation of the average mass loss (%) with the material lling ratio.
S.S. olak et al. / Powder Technology 325 (2018) 678686 685
Table 18
The results of the tumbling tests.
Clinker Mass of ball charge (g) 19,902.66 19,894 19,888.88 19,887.53 19,886.57
80% passing size (micron) 44.23 32.88 28.13 24.36
Slag Mass of ball charge (g) 19,936.77 19,927.35 19,922.17 19,911.58 19,911.08
80% passing size (micron) 77.18 71.62 56.01 26.55
Copper ore Mass of ball charge (g) 19,956.27 19,947.01 19,941.23 19,939.19 19,936.77
80% passing size (micron) 46.31 43.41 29.87 25.64
1.80 taking the carbon, silicon and phosphorus elements, which are affecting
Slag
1.60
Cumulative media wear
1.00 P% Si%
CEV C% 2
0.80 3
0.60
0.40
The variation of the content of primary carbide with the carbon
0.20
equivalent value is depicted in Fig. 17. As can be seen, these two param-
0.00
eters are directly proportional to each other. Additionally, it should be
0 100 200 300 400 500
Tumbling time (minutes)
emphasized that the A1 type of media, which has the highest rate of
wear, has the lowest CEV while B2 type of media, which has the least
Fig. 15. The variation of the wear rate with the type of material used.
rate of wear, has the second highest CEV value.
In order to investigate the effect of primary carbides on the wear
rates, B1 and A1 type of media having similar amounts of the secondary
carbides were compared. The results of the tests showed that B1 type of
carbides with the elements were investigated. Initially, the relationship media had higher wear rate than that of A1 type. When the chemical
between the amount of molybdenum and secondary carbides were compositions of these two are considered, it is found out that the in-
studied and illustrated in Fig. 16. It should be emphasized that only crease in the amount of carbon increases the formation of the primary
A2, A3, B1 and B2 types of media includes molybdenum. carbides and the rate of wear decreases.
As can be seen from Fig. 16, there is a strong correlation between the As a conclusion of the evaluations, it is understood that, the forma-
volumetric fraction of secondary carbides and the molybdenum tion of martensite, primary and secondary carbides depend on the
content. chemical composition and affect the wear rates of the media directly.
The formation of the primary carbides also depends on the chemical
composition of the media. In this context, Brown [18] derived an 3.5. The correlation between the laboratory and industrial scale test results
equation (Eq. 2) and calculated carbon equivalent values (CEV) by
The aim of this work is to develop a media wear testing procedure
that could be utilized by both media and cement manufacturers. Conse-
quently, it is important to investigate the relationship between the lab-
oratory and industrial scale results. Within the scope, long-term
observations (6 months) were made at 3 different cement plants during
CEM I 42.5R production where one of the tested grinding balls was
charged (A1 type of media). At the same time, the material processed
by those cement manufacturers were utilized in the laboratory and
wear measurements for 900 min were undertaken. The calculations of
the wear rate were made based on the mass loss in the ball charge per
ton of material milled (g/t) and illustrated in Fig. 18.
Fig. 16. The variation of the amount of secondary carbides with molybdenum content.
Table 19
The variation of the mass loss with martensite and secondary carbides.
A1 60% 6% 0.084%
A3 15% 0.050%
A2 50% 5% 0.070%
C1 10% 0.047%
B1 40% 6% 0.064% Fig. 17. The relationship between the carbon equivalent value and the amount of the
B2 12% 0.033% primary carbides.
686 S.S. olak et al. / Powder Technology 325 (2018) 678686
Within the scope of the study, the microstructures of the media were
investigated and correlations were developed with the wear rates of the
media. As a conclusion of the evaluations, it was found out that, the
formation of martensite, primary and secondary carbides depended on
the chemical composition and affected the wear rates of the media
directly. It was understood from the studies that the higher the
martensite content the higher the mass loss of the media. It was also
concluded that the increase in the amount of both primary and second-
ary carbides resulted in reduction in the wear rate of the media.
The signicant ndings of this study are summarized below;
1. The methodology is sensitive to the changes in the milling conditions.
2. No matter whether the material is inside of the mill or what size of
the mill selected for the testing purpose, it was calculated that the
media from the highest wear rate to the lowest was in the order of
A1 N A2 N B1 N A3 N C1 N B2. This implies that the testing methodol-
Fig. 18. Comparison between the laboratory and industrial scale results. ogy can produce comparative results hence a better understanding of
the media properties can be provided and the media quality or the
economy of the milling operation can be improved.
As can be seen from Fig. 18, the laboratory results are in good agree- 3. There is a strong correlation between the micro-structure and the
ment with the industrial data. In other words, as a result of the test work wear rate of the media.
carried out at the laboratory scale, the wear rates in the industrial scale 4. There is a correlation between the laboratory scale wear rate and in-
can be predicted. dustrial scale ones hence the operational wear rates can be predicted
by applying the procedure dened in this paper.
4. Conclusions
Acknowledgements
This study focussed on developing a test procedure measuring the
wear rate of the media used in the cement grinding applications. So The authors would like to thank ema Dkm Sanayi for providing
far, there have been studies conducted for the wet milling applications different types of media and Hacettepe University Mining Engineering
however for the dry milling there is a need for a research to contribute Department for providing the laboratory facilities.
the savings of the overall process. Within the scope of the study, six
types of media were subjected to the tumbling tests at different milling References
conditions i.e., time, material type, material lling, media lling, mill [1] L.A. Vermeulen, D.D. Howat, C.L.A.I.M.M. Gought, Theories of Ball Wear and the Re-
geometries, with the aim of revealing the differences between the sults of a Marked-Ball Test in Ball Milling, 1983.
[2] C. Aldrich, Consumption of steel grinding media in mills a review, Miner. Eng. 49
types of media. (2013) 7791.
The test works without material inside the mill were performed at 3 [3] T.W. Chenje, D.J. Simbi, E. Navara, Relationship between microstructure, hardness,
different mill dimensions. The results indicated that the wear rate of the impact toughness and wear performance of selected grinding media for mineral
ore milling operations, Mater. Des. 25 (1) (2004) 1118.
media increased with the increase in mill diameter. Although different [4] R. Hukki, Correlation between principal parameters affecting mechanical ball wear,
results were obtained from different mill geometries, the order from Trans. AIME 199 (1954) 642644.
[5] I. Iwasaki, R.L. Pozzo, K.A. Natarajan, K. Adam, J.N. Orlich, Nature of corrosive and
the highest to the lowest wear were as; A1 N A2 N B1 N A3 N C1 N B2, abrasive wear in ball mill grinding, Int. J. Miner. Process. 22 (14) (Apr. 1988)
independent of the mill dimensions. In other words, A1 type of media 345360.
had the highest rate of wear where the B2 type of media had the lowest [6] C.P. Massola, A.P. Chaves, E. Albertin, A discussion on the measurement of grinding
media wear, J. Mater. Res. Technol. 5 (3) (2016) 282288.
rate for each of the mill dimension. [7] M.K. Yelloji Rao, K.A. Natarajan, Effect of electrochemical interactions among sul-
A series of test studies were undertaken when the material was phide minerals and grinding medium on the otation of sphalerite and galena,
inside the mill. Briey, the scope was Int. J. Miner. Process. 29 (34) (Oct. 1990) 175194.
[8] H. Dndar, Investigation of the Grinding Parameters of the Steel Slags Cooled at
Different Cooling Rates, 2006.
Measuring the wear rate of the media at different material lling,
[9] I.E. Rivera Madrid, B.. Rodrguez, O. Bustamante, O.J.R. Baena, J.M. Menndez-
Measuring the wear rates of different types of media for the same Aguado, Ceramic ball wear prediction in tumbling mills as a grinding media selec-
product, tion tool, Powder Technol. 268 (2014) 373376.
[10] F. Bond, Crushing and Grinding CalculationsAllis-Chalmers Manufacturing Company
Measuring the wear rates of the media for different types of material. 1961.
[11] J.D. Gates, A. Giblett, R. Dunne, Prediction of Plant Ball Mill Media Wear Rates from
Laboratory Ball Mill Test Data, 2012 123.
[12] A. Mular, G. Jergensen, Design and Installation of Comminution Circuits, 1982.
The results concluded that the wear rate decreased signicantly [13] J. M. deS, Applied Statistics: Using SPSS, STATISTICA, and MATLAB, 2003.
with the increase in the material load, which could be attributed to [14] D.D. Wackerly, W. Mendenhall, R.L. Scheaffer, Mathematical Statistics with
Applications, Thomson Brooks/Cole, 2008.
the reduced ball to ball interaction since the material had cushioning [15] Q. Liu, Microstructure Evaluation and Wear-Resistant Properties of Ti-Alloyed
effect on the surfaces of the media. The tests for the same material Hypereutectic High Chromium Cast Iron, 2013.
[16] E. Zdraveck, J. Tkov, M. Ond, Effect of microstructure factors on abrasion
showed that the media from the highest wear rate to the lowest was resistance of high-strength steels, Res. Agric. Eng. 60 (3) (2014) 115120.
in the order of A1 N A2 N B1 N A3 N C1 N B2 that was similar to that of [17] A. Kootsookos, J.D. Gates, The role of secondary carbide precipitation on the fracture
the tests without the material. Finally, it was observed that different toughness of a reduced carbon white iron, Mater. Sci. Eng. A 490 (12) (2008)
313318.
types of materials resulted in obtaining different wear rates that could [18] J. Brown, Foseco Ferrous Foundryman Handbook, p. 76, 2000.
be the result of having different grindability values.