- Email: [email protected]
Effects of mill design and process parameters in milling dry extrudates
Powder Technology 278 (2015) 84–93
Contents lists available at ScienceDirect
Powder Technology journal homepage: www.elsevier.com/locate/powtec
Effects of mill design and process parameters in milling dry extrudates Aditya Vanarase a, Rizwan Aslam b, Sarang Oka b, Fernando Muzzio b,⁎ a b
Bristol-Meyers Squibb, 1 Squibb Drive, New Brunswick, NJ 08903, United States Department of Chemical & Biochemical Engineering, Rutgers, The State University of New Jersey, NJ, United States
a r t i c l e
i n f o
Article history: Received 30 November 2014 Received in revised form 29 January 2015 Accepted 13 February 2015 Available online 20 February 2015 Keywords: Milling Cone mill Hammer mill Process parameters Particle size distribution
a b s t r a c t An experimental study was performed to characterize two continuous mills for their ability to mill alumina–magnesia extrudates. The effect of mill parameters, namely, the screen aperture size, and impeller speed on the particle size distribution of the milled product was quantified for a conical screen mill and a hammer mill. In general, the conical screen mill was found to be more sensitive to changes in impeller speed compared to the hammer mill. The effect of impeller speed in case of the hammer mill was non-monotonic while the increasing speeds led to reduction in particle size in case of the cone mill, for the same screen aperture size. The effect of aperture screen size was observed to play a dominant role in dictating particle size distribution of the product material for both mills. In case of the cone mill, grated type screens exhibited higher milling capacity than round screens with equivalent apertures. Lastly, a study comparing the statistical particle size distribution parameters was performed for process design purposes. It was deduced that, if the desired particle size is greater, the comil provides a narrower particle size distributions than the hammer mill; whereas if the desired particle size is smaller, both mills exhibit similar poly-dispersity. The study provided insight into fundamental breakage mechanisms for both mill classes. Breakage in the hammer mill occurs primarily due to the impact of the hammers and large particles may often leak through the mill without sufficient breakage. Breakage in the comil is more gradual as the impeller sweeps a wide area generally ensuring sufficient breakage of particles before they exit the milling chamber. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Granular milling processes are used throughout industry. For example, in the pharmaceutical industry, milling is used at various manufacturing stages, including wet and/or dry milling after crystallization of drugs to reduce the particle size or achieve the desired PSD [1], de-lumping of dry powder ingredients to de-agglomerate and improve power handling, milling of ribbons after roller compaction, and delumping of the wet mass in wet granulation [2,3]. Other industries make similar or even more extensive use of milling steps. For example, in catalyst manufacturing, ingredients are often de-lumped, wet granulated, extruded, dried, and milled again. Similar sequences involving multiple milling steps are common in food processing, cosmetics manufacturing, minerals processing, etc. In order to integrate mills effectively into continuous manufacturing processes, developing predictive understanding of continuous mills is necessary. This particular study is aimed at understanding the size reduction behavior in continuous mills. Alumina–magnesia dry extrudates, a ceramic material used in the manufacture of catalyst
⁎ Corresponding author. Tel.: +1 848 445 335. E-mail address: [email protected] (F. Muzzio).
http://dx.doi.org/10.1016/j.powtec.2015.02.021 0032-5910/© 2015 Elsevier B.V. All rights reserved.
supports, was used as the model material. In this study, a methodology to optimize the particle size distribution under process constraints such as milling capacity and/or PSD requirements is presented. Often times in the manufacturing environment, the processing rate of the milling operation needs to be changed depending on the upstream/downstream process requirements, while maintaining the PSD of the milled material within the desired specifications. In such cases, the material response to operational parameters of the mill needs to be well understood. Typical industrial continuous mills include hammer mills, conical mills, pin mills, knife mills and reciprocating mills. Population balance models haven been applied for modeling the milling performance in hammer mills [3,4] and conical mills [5]. Hammer mills can also be modeled based on the damage mechanics of the material [6,7]. The population balance models built on the batch milling case have been applied for the equivalent continuous mill by assuming perfect mixing conditions and measuring residence times [8,9]. However, studies comparing the performance of different mills for the same materials remain scarce, and quantitative methodologies for comparing the resulting PSDs need further development. In this paper, we attempt to address both of these gaps. This study is focused on a comparison between two commonly used continuous mills—a conical mill and a hammer mill. The relative effects of
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
85
2. Equipment
material used in this study is very brittle and undergoes fragmentation (resembling many dry granulated and some wet granulated materials following drying). In order to impose the least amount of shear on the material, an impeller with a round cross section (Fig. 2(b)) was used.
2.1. Conical mill (Comil—Quadro)
2.2. Hammer mill (Fitzpatrick)
The conical mill used in this study was a Quadro Comil Model197-S (Fig. 1(a)). Following recommendations of the equipment manufacturer, the conical impeller was positioned as close as possible to the screen (Fig. 1(b)). Two types of screens (Fig. 2(a)), one with round holes, and another with a combination of round and grated holes, were used in this study. The Comil operates by applying intense shear in a narrow region between an impeller and the screen, which creates frictional contacts and drives particles through the perforations of the screen. The choice of screen is very important because the perforation size and spacing greatly affect the PSD. The
The hammer mill used in this study is manufactured by Fitzpatrick (Fig. 3(a)). The blades of the hammer mill pulverize particles along their rotational path. The shape of hammers is shown in Fig. 3(d). As shown in Fig. 1(b), the tips of the blades create a narrow region of intense shear near the surface of the screen. The screen size in this setup has less importance because the shearing is mostly being done by the blades. The screens used in this mill are of a ‘C’ shape (Fig. 3(c)).
operational parameters are examined and quantitative criteria for comparing PSDs are introduced.
3. Materials The material used in this study, called ‘Material 1’, is a mixture of aluminum and magnesium oxides that has been mixed as a wet paste and then calcined at high temperatures, and is used in the manufacture of catalyst supports. The initial material was in the form of extrudate pellets as shown in Fig. 4(a). The pellets were 0.5 cm in diameter and variable in lengths. The length varied between ~ 0.5 and 3 cm. 4. Particle size distribution measurement method A vibrational sieve shaker, ‘Endcotts Octagon 2000’, shown in Fig. 5, was used for sieving. The milled powder was sieved into 9 size bins. The sieves had 8″ diameter and U.S. standard sieve numbers of 10, 14, 18, 30, 40, 60, 120 and 230 were used in the PSD measurement. The sieving procedure used can be described as follows. All the trays were cleaned and dried before the sieve analysis. A sample of 200 g of material was loaded on the top sieve. The sieve tray assembly was then placed on the shaker table and secured with the shaker tables clamp. The shaker table was set to power 8, and run for 15 min. After sieving, the mass of the powder on individual trays was weighed. A bar chart (Fig. 6(a)) of mass fraction vs. size bins was thus created. Particle diameters (d10, d50, or d90) were read from the cumulative PSD graph using linear interpolation. 5. Experimental conditions The experimental conditions examined for the Comil and hammer mill are presented in Tables 1 and 2 respectively. The range of impeller speeds used in both mills was similar. The milling chamber of hammer mill is larger than that of the Comil, which makes the tip speeds in the hammer mill greater than those in the Comil for equivalent impeller RPMs. 6. Results 6.1. Effect of operational parameters on the flow rate and impeller passes in the Comil
Fig. 1. (a) Conical mill (Comil—Quadro Model # 197); (b) milling chamber with conical round impeller.
The PSD of the milled material is dependent on its residence time in the mill. The residence time of the material in the mill was estimated by measuring the material hold-up in the mill and dividing it by the mass flow rate. Out of the two possible feeding modes (chocked feeding and starved feeding), the Comil was operated under chocked feeding condition. Under the starved feeding condition, flow rate of the incoming material is constant, and hold-up (mass) in mill varies as a function of the operational parameters (screen, impeller speed); whereas under the chocked feeding
86
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 2. (a) Comil screen designs, round holes (R) vs. round-grated holes (RG); (b) impeller design.
condition, the hold-up of the material in the mill is nearly constant, and flow rate is dictated by the operational parameters. The choice between ‘starved’ and ‘choked’ operation is often dictated by process constraints and desired product attributes. For continuous processes, operating at specified mass-rates, milling operations invariably occur under starved conditions, unless, the mass-rate overwhelms the mill. The mill parameters are tuned to achieve the desired product properties. Operating under choked condition enables one to exert more control on the total energy input to the material. Operations in the choked mode are also typified by reduced dusting and fewer fines. The effect of operational parameters on flow rate is shown in Fig. 7(a). Some variations in the hold-up are expected because the PSD of the product in the milling chamber varies as a function of the operational parameters of the mill. These variations in this case are relatively small since the volume of the milling chamber is smaller than the total empty volume in the mill, which includes the milling chamber as well as the feeding section. The residence time (hold-up/flow rate) of the material in the mill was determined by measuring the hold-up and flow rate under steady state. As depicted in Fig. 7(a), the flow rate was found to increase linearly with an increase in the impeller speed. The screen hole size and the screen geometry significantly affected the flow rate. As anticipated, screens with bigger hole-diameters showed higher flow rates. A screen with a combination of round and grated type holes exhibited higher flow rate than the round holed screen with equivalent diameter. Residence time measurement can be used to estimate the total number of blade passes. The number of blade passes applied on the material
during its residence time in the mill was calculated using the relationship given in Eq. (1). Number of impeller passes ðRevsÞ Hold‐up ðkgÞ ¼ Impeller Speed ðRevs=sÞ Flow rate ðkg=sÞ
ð1Þ
The behavior of the number of impeller passes as a function of operational parameters is shown in Fig. 7(b). As expected, the number of impeller passes exhibited the following order—991 μm R N 1575 μm R N 1905 μm R N 1575 μm RG. However, and less expected, the relationship between the impeller speed and impeller passes was strongly affected by the screen size/type. The number of impeller passes was the maximum at the intermediate impeller speed for 991 μm R screen, whereas it decreased with increase in impeller speed for 1575 μm RG screen. Other screens, 1575 μm R and 1905 μm R, showed a relatively constant number of impeller passes, and a slight decreasing trend as a function of impeller speed respectively. These results show that the relationships between impeller speed and impeller passes are non-monotonic, and they need to be characterized individually. With an increase in impeller speed, the impact force exerted by the impeller while it passes through the material bed increases, which essentially decreases the particle size. With increase in the number of passes, the total shear applied on the material increase which also decreases particle size. These observations can be used to interpret the size reduction behavior observed in the mill, and draw conclusions about the underlying milling mechanisms.
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
87
Fig. 3. Hammer mill (Fitzpatrick, Model—DASO6). (a) Hammer mill equipment, (b) position of the impeller and screen, (c) screen design and (d) hammers.
6.2. Effect of operational parameters on PSD The effect of operational parameters on the PSD was analyzed by tracking changes in certain statistical PSD parameters (d10, d50 and d90). A comparison between the two mills (Comil and hammer mill) is shown in Fig. 8.
6.2.1. Effect of screen size In both mills, with an increase in the screen diameter, particle size at the mill discharge increases. For the Comil, as shown previously, an increase in the screen diameter decreases the residence time in the mill, meaning fewer number of impeller passes (breaking action) on the material. A correlation between impeller passes and particle size is shown in Fig. 9. For screens with the same hole shape (R), an increase in the number of impeller passes leads to decrease in particle size. The trend is more visible for particles of greater size. The monotonic relationship between the number of impeller passes and particle size does not hold for screens with different hole shapes. As shown in Fig. 9, for RG type of screens, although fewer number of impeller passes are exerted, an equivalent increase in particle size is not observed.
Qualitatively, the effect of screen diameter in the hammer mill is similar to that of the Comil; however the effect of residence time could not be tested. Residence time measurements for the hammer mill were not available because of the difficulty in measuring hold-up and flow rate under controlled conditions. The effectiveness of the screen was analyzed by calculating α values as a function of the operational parameters (Fig. 10). The definition of α is given in Eq. (2).
α¼
d95 Screen Aperture
ð2Þ
In the Comil, the RG screen provides greater α values than R screens. The RG screen, because of the inherent shape, provides less resistance to the material flow, which increases particle size. In the case of the R screens, the differences between the α values were minor; the 991 μm screen showed slightly greater values. Further comparison between screens is difficult using this data since the screens are also different with respect to the % open area and thickness, and not just the aperture. α values were found to be higher for the hammer mill than the Comil.
88
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 4. Material (a) before and (b) after milling.
6.2.2. Effect of the mill and the screen size on the relative effect of impeller speed (REIS) The effect of impeller speed in the two mills was clarified by analyzing the sensitivity of the particle size distribution to impeller speed, defined as the rate of change of statistical parameters (d10, d25, d50, d75 and d90) as a function of the impeller speed. A linear regression analysis was performed on a dataset of particle size vs. speed at each size cut. The quantity plotted on the Y-axis in Fig. 11 is the slope of this linear fit, normalized by the average particle size. This new quantity was defined as the relative effect of impeller speed (REIS). The equation below was used to compute the REIS values for various screens. Change in dx Change in impeller speed REIS ¼ d50
ð3Þ
The (REIS) value was plotted against the size cut for each screen. The quantity captures the sensitivity of a particle size distribution parameter with changing impeller speed. The larger the modulus of the value, the greater is the impact of the impeller speed. The value assists in making judicious design choices for choosing impeller speeds when constrained by mill capacities and desired particle size distributions. As shown in Fig. 11(a), for the Comil, the relative effect of impeller speed decreases (less negative slope) with increase in size cut. This trend is observed between d25 and d90. For the Comil, a comparison between the screens was made using the REIS values. For the round holed screens, an increase in screen diameter from 991 to 1575 microns increases the REIS. A further increase in the screen diameter decreases the REIS. For the RG type screen, even lower rates were observed. Thus the REIS seems to be a function of
Fig. 5. Sieve shaker.
both the residence time in the mill, and the inherent size reduction rate which is a function of the initial particle size. Typically, the size reduction rate decreases with decrease in particle size, and the residence time in the mill increases with decrease in the screen diameter (except for different screen geometry)—991R N 1575R N 1905R N 1575G. With these two phenomena coupled together, the REIS increases only until the screen diameter reaches 1575 microns; any further increase in screen diameter (decrease in residence time) makes the particles pass through the screen without any significant breaking action. A completely different behavior was observed for the hammer mill. The REIS was found to decrease with increase in particle size and then increase again. The absolute REIS values were much smaller than the Comil for equivalent screen sizes. This behavior suggests that in the hammer mill, the larger particles exit through the screen at a relatively faster rate. Air entrainment in the mill might be playing a role in this case. The hammer mill was operated under the starved feeding condition, as opposed to the chocked feeding condition in the case of the Comil. Once the extrudate material enters the mill, it is broken down into several fragments. These fragments settle on the screen, and are impacted by the hammers at every revolution. While the hammer mill is operating, significant amount of air is entrained, which causes the number of contacts between hammer and particles or particle–particle to decrease. Further size reduction of the initially created fragments happens primarily though abrasion. During this process, once the particles are reduced to a size sufficiently smaller than the screen size, they make an
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
89
Table 2 Experimental conditions for Hammer mill. Agitation rates (RPM)
1419.17 2240.83 3062.5
Screen type (diameter in μm) 1000 R
1500 R
2000 R
X X X
X X X
X X X
Tip speeds (cm/s)
1462.75 2309.64 3156.53
correspond to different screens. The solid markers represent data points corresponding to the Comil, and the open makers represent the data points corresponding to the hammer mill. As shown in Fig. 12(a), for the case of 991 μm screen, d10 does not show any particular trend with respect to d50. However for other screens an increase in d10 was observed as a function of increase in d50. When all the data points corresponding to different screens and speeds are considered together, a plateau in the d10 as a function of d50 was observed. For the case of hammer mill, for individual screens, a clear trend between d10 and d50 was not observed. However, when all the data points are considered together, a relatively linear relationship with a relatively high degree of correlation (R2 = 0.83) was observed. The hammer mill shows lower d10 values (i.e., a broader PSD) than the Comil at equivalent d50. The behavior of d90 as a function of d50 is shown in Fig. 12(b). For the case of the Comil, two linear regions can be identified. The data-points for the 991 μm screen lie on a separate linear curve than
Fig. 6. Calculation of PSD parameters using mass fractions collected on sieves. (a) Raw data of mass fraction vs. size bins; (b) calculation of PSD parameters by linear interpolation.
exit (leak) through the screen. This phenomenon creates a significant amount of fines. This could be a possible milling mechanism in the hammer mill. In the Comil, the number of contacts (both particle–particle and particle–hammer) per unit volume and per time is significantly greater than the hammer mill. This makes the size reduction process more continuous (gradual) as opposed to the hammer mill. The leakage of bigger particles through the screen, which was observed for the hammer mill, is less likely for the Comil, since particles have to pass through the material bed in the mill, and also through the shear zone between impeller and screen.
6.3. Correlations between the statistical particle size parameters The PSD data from the milling experiments can be used for design purposes. In this analysis, in Fig. 12, the d10, d25, d75 and d90 values are plotted against the d50 of the particle size distribution from the individual milling experiments. The d50 of the particle size distribution is used as the design variable. The different markers in the figure
Table 1 Experimental conditions for Comil (‘R’—Round holes, ‘RG’—combination of round and grated holes). Agitation rates (RPM)
1419.17 2281.8 3144.41
Screen type (diameter in μm)
Tip speeds (cm/s)
991 R
1575 R
1575 RG
1905 R
X X X
X X X
X X X
X X X
802.15 1289.77 1777.30
Fig. 7. (a) Effect of Comil impeller speed and screen size on the flow rate; (b) effect of Comil screen size and impeller speed on impeller passes.
90
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 8. Effect of impeller tip speed on the PSD. (a) d10 vs. tip speed, (b) d50 vs. tip speed and (c) d90 vs. tip speed for both mills. Different level curves represent different screens.
the rest of the two screens. However for the hammer mill, a single linear relationship (R2 = 0.99) was observed when all the data points are considered together. The hammer mill shows higher d90 than that of the Comil for equivalent d50 values. Other statistical parameters including d75 and d25 exhibited fairly good linear correlations with d50 for both the mills (Comil: d25—R2 = 0.92, d75 R2 = 0.97, and Hammer mill: d25—R2 = 0.99, d75 R2 = 0.97). Hammer mill shows lower d25 values, and higher d75 values than those for the Comil, again indicating a broader PSD. These results show that milling in a hammer mill results into larger coarse particles (d75 and d90), and smaller fine particles (d10 and d25). For the case of the Comil, the opposite behavior was observed. It produces larger fine particles (d10 and d25), and smaller coarse particles (d75 and d90). The Comil thus produces a narrower PSD. Using this methodology, the desired PSD is selected by comparing one statistical parameter against every other parameter. Once the desired PSD is decided, operating conditions can further be optimized based on the throughput requirements or energy constraints.
the PSD, and normalizing it by the median (d50). For the hammer mill (Fig. 13(b)), the PDI was nearly constant and it did not vary significantly with d50. In the case of the Comil (Fig. 13(a)) the PDI showed three distinct linear relationships depending on the screen size/type used. In the first relationship, the data points corresponding to 1575 μm and 1905 μm round holed screens lie on the same linear curve. The second curve and the third curve correspond to the 1575 μm RG screen and the 991 μm R screen respectively. In all the three curves an increase in the impeller speeds leads to decrease in the PDI. These results show that Comil produces narrower particle size distributions when the desired particle size is large. Both mills produce nearly the same particle size distribution when the desired particle size is smaller. It seems that the PSDs are not scalable as a function of either impeller speed, or screen size. A combination of both the parameters needs to be used in order to scale the PSDs. For the Comil, for screens with similar geometry, if a higher particle size is desired, larger screen and smaller speed are desired. For the hammer mill, the effect d50 on the PDI is different for different screens. It is difficult to deduce any scaling rule using this dataset for the hammer mill.
6.4. Scaling behavior of PSDs 7. Conclusions The scaling behavior of PSDs as a function of operational parameters was analyzed by using the poly-dispersity index (PDI). The PDI was defined by taking the ratio of the difference between the d90 and d10 of
In this study, size reduction behavior of two types of continuous mills, a conical mill—Quadro-Comil, and a hammer mill—Fitzpatrick,
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
91
Fig. 9. Correlation between impeller passes particle size. (a) d10 vs. impeller passes, (b) d50 vs. impeller passes and (c) d90 vs, impeller passes for the Comil.
was characterized for alumina-magnesia extrudates. The effects of operational variables (screen size, shape and impeller speed) on the PSD of milled product have been summarized below. Although, the effect of
Fig. 10. Effectiveness of screen for the Comil and hammer mill.
mill parameters on product properties is a strong function of the material attributes of the feed, the study sheds light on the breakage mechanism in the two mills and consequently, the subtle differences in the effect of milling parameters on process performance between the two mills. Qualitatively the effect of screen size was similar in both mills; an increase in screen size leads to increase in particle size. In the Comil, for screens with identical shape of the aperture, an increase in the number of impeller passes leads to a decrease in particle size. The grated type screens exhibited higher milling capacity than round screens with equivalent apertures. The effect of speed was different in the two mills. For the Comil, except for fines (d10), generally, an increase in impeller speed lead to a decrease in particle size. Smaller particles showed higher sensitivity to impeller speed. For the hammer mill, the nonmonotonic effects of speed were observed across different size cuts in the PSD. Overall the sensitivity of impeller speed to size reduction was less in the hammer mill, and the extrema of the PSD parameters showed higher sensitivity to impeller speed. A comparison between the statistical PSD parameters was performed for process design purposes. If the desired particle size is greater, the Comil shows narrower particle size distributions than the hammer mill; whereas if the desired particle size is smaller, both mills show similar poly-dispersity. In the hammer mill greater amounts of fines are created primarily because of the abrasion of the large particles. Large particles tend to leak through the screen, often without sufficient breaking action. In the Comil, a more gradual size reduction process is
92
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 11. Sensitivity of particle size to impeller speed in (a) Comil and (b) Hammer mill.
Fig. 12. Correlations between the statistical parameters of PSD. (a) d10 vs. d50, (b) d90 vs. d50, (c) d25 vs. d50 and (d) d75 vs. d50.
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
93
observed. The creation of fines is primarily through the impact of hammers. The leakage of large particles through the screen is relatively less since the swept area by the impeller against the screen in the Comil is greater than the hammer mill. This study provides design rules for the milling process, under capacity and constraints of PSD parameters. Also a methodology to select optimal combination of operational variables, for a given particle size is presented. References [1] Y. Chen, Y. Ding, D.G. Papadopoulos, M. Ghadiri, Energy-based analysis of milling αlactose monohydrate, J. Pharm. Sci. 93 (2004) 886–895. [2] L.R. Schenck, R.V. Plank, Impact milling of pharmaceutical agglomerates in the wet and dry states, Int. J. Pharm. 348 (2008) 18–26. [3] J.J.A.M. Verheezen, K. van der Voort Maarschalk, F. Faassen, H. Vromans, Milling of agglomerates in an impact mill, Int. J. Pharm. 278 (2004) 165–172. [4] F. Shi, T. Kojovic, J.S. Esterle, D. David, An energy-based model for swing hammer mills, Int. J. Miner. Process. 71 (2003) 147–166. [5] G.K. Reynolds, Modelling of pharmaceutical granule size reduction in a conical screen mill, Chem. Eng. J. 164 (2010) 383–392. [6] L.G. Austin, R.R. Klimpel, P.T. Luckie, Process Engineering of Size Reduction: Ball Milling, Society of Mining Engineers of the American Institute of Mining, Metallurgical, and Petroleum Engineering, Inc., New York, 1984. 22–28. [7] L.G. Austin, A preliminary simulation model for fine grinding in high speed hammer mills, Powder Technol. 143–144 (2004) 240–252. [8] H. Berthiaux, C. Chiron, J. Dodds, Modelling fine grinding in a fluidized bed opposed jet mill: Part II: Continuous grinding, Powder Technol. 106 (1999) 88. [9] H. Berthiaux, J. Dodds, Modelling fine grinding in a fluidized bed opposed jet mill: Part I: Batch grinding kinetics, Powder Technol. 106 (1999) 78.
Fig. 13. Effect of operational parameters on the scaling behavior of PSDs. (a) Polydispersity index vs. median (d50) for the Comil; (b) poly-dispersity index vs. median (d50) for hammer mill.
Contents lists available at ScienceDirect
Powder Technology journal homepage: www.elsevier.com/locate/powtec
Effects of mill design and process parameters in milling dry extrudates Aditya Vanarase a, Rizwan Aslam b, Sarang Oka b, Fernando Muzzio b,⁎ a b
Bristol-Meyers Squibb, 1 Squibb Drive, New Brunswick, NJ 08903, United States Department of Chemical & Biochemical Engineering, Rutgers, The State University of New Jersey, NJ, United States
a r t i c l e
i n f o
Article history: Received 30 November 2014 Received in revised form 29 January 2015 Accepted 13 February 2015 Available online 20 February 2015 Keywords: Milling Cone mill Hammer mill Process parameters Particle size distribution
a b s t r a c t An experimental study was performed to characterize two continuous mills for their ability to mill alumina–magnesia extrudates. The effect of mill parameters, namely, the screen aperture size, and impeller speed on the particle size distribution of the milled product was quantified for a conical screen mill and a hammer mill. In general, the conical screen mill was found to be more sensitive to changes in impeller speed compared to the hammer mill. The effect of impeller speed in case of the hammer mill was non-monotonic while the increasing speeds led to reduction in particle size in case of the cone mill, for the same screen aperture size. The effect of aperture screen size was observed to play a dominant role in dictating particle size distribution of the product material for both mills. In case of the cone mill, grated type screens exhibited higher milling capacity than round screens with equivalent apertures. Lastly, a study comparing the statistical particle size distribution parameters was performed for process design purposes. It was deduced that, if the desired particle size is greater, the comil provides a narrower particle size distributions than the hammer mill; whereas if the desired particle size is smaller, both mills exhibit similar poly-dispersity. The study provided insight into fundamental breakage mechanisms for both mill classes. Breakage in the hammer mill occurs primarily due to the impact of the hammers and large particles may often leak through the mill without sufficient breakage. Breakage in the comil is more gradual as the impeller sweeps a wide area generally ensuring sufficient breakage of particles before they exit the milling chamber. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Granular milling processes are used throughout industry. For example, in the pharmaceutical industry, milling is used at various manufacturing stages, including wet and/or dry milling after crystallization of drugs to reduce the particle size or achieve the desired PSD [1], de-lumping of dry powder ingredients to de-agglomerate and improve power handling, milling of ribbons after roller compaction, and delumping of the wet mass in wet granulation [2,3]. Other industries make similar or even more extensive use of milling steps. For example, in catalyst manufacturing, ingredients are often de-lumped, wet granulated, extruded, dried, and milled again. Similar sequences involving multiple milling steps are common in food processing, cosmetics manufacturing, minerals processing, etc. In order to integrate mills effectively into continuous manufacturing processes, developing predictive understanding of continuous mills is necessary. This particular study is aimed at understanding the size reduction behavior in continuous mills. Alumina–magnesia dry extrudates, a ceramic material used in the manufacture of catalyst
⁎ Corresponding author. Tel.: +1 848 445 335. E-mail address: [email protected] (F. Muzzio).
http://dx.doi.org/10.1016/j.powtec.2015.02.021 0032-5910/© 2015 Elsevier B.V. All rights reserved.
supports, was used as the model material. In this study, a methodology to optimize the particle size distribution under process constraints such as milling capacity and/or PSD requirements is presented. Often times in the manufacturing environment, the processing rate of the milling operation needs to be changed depending on the upstream/downstream process requirements, while maintaining the PSD of the milled material within the desired specifications. In such cases, the material response to operational parameters of the mill needs to be well understood. Typical industrial continuous mills include hammer mills, conical mills, pin mills, knife mills and reciprocating mills. Population balance models haven been applied for modeling the milling performance in hammer mills [3,4] and conical mills [5]. Hammer mills can also be modeled based on the damage mechanics of the material [6,7]. The population balance models built on the batch milling case have been applied for the equivalent continuous mill by assuming perfect mixing conditions and measuring residence times [8,9]. However, studies comparing the performance of different mills for the same materials remain scarce, and quantitative methodologies for comparing the resulting PSDs need further development. In this paper, we attempt to address both of these gaps. This study is focused on a comparison between two commonly used continuous mills—a conical mill and a hammer mill. The relative effects of
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
85
2. Equipment
material used in this study is very brittle and undergoes fragmentation (resembling many dry granulated and some wet granulated materials following drying). In order to impose the least amount of shear on the material, an impeller with a round cross section (Fig. 2(b)) was used.
2.1. Conical mill (Comil—Quadro)
2.2. Hammer mill (Fitzpatrick)
The conical mill used in this study was a Quadro Comil Model197-S (Fig. 1(a)). Following recommendations of the equipment manufacturer, the conical impeller was positioned as close as possible to the screen (Fig. 1(b)). Two types of screens (Fig. 2(a)), one with round holes, and another with a combination of round and grated holes, were used in this study. The Comil operates by applying intense shear in a narrow region between an impeller and the screen, which creates frictional contacts and drives particles through the perforations of the screen. The choice of screen is very important because the perforation size and spacing greatly affect the PSD. The
The hammer mill used in this study is manufactured by Fitzpatrick (Fig. 3(a)). The blades of the hammer mill pulverize particles along their rotational path. The shape of hammers is shown in Fig. 3(d). As shown in Fig. 1(b), the tips of the blades create a narrow region of intense shear near the surface of the screen. The screen size in this setup has less importance because the shearing is mostly being done by the blades. The screens used in this mill are of a ‘C’ shape (Fig. 3(c)).
operational parameters are examined and quantitative criteria for comparing PSDs are introduced.
3. Materials The material used in this study, called ‘Material 1’, is a mixture of aluminum and magnesium oxides that has been mixed as a wet paste and then calcined at high temperatures, and is used in the manufacture of catalyst supports. The initial material was in the form of extrudate pellets as shown in Fig. 4(a). The pellets were 0.5 cm in diameter and variable in lengths. The length varied between ~ 0.5 and 3 cm. 4. Particle size distribution measurement method A vibrational sieve shaker, ‘Endcotts Octagon 2000’, shown in Fig. 5, was used for sieving. The milled powder was sieved into 9 size bins. The sieves had 8″ diameter and U.S. standard sieve numbers of 10, 14, 18, 30, 40, 60, 120 and 230 were used in the PSD measurement. The sieving procedure used can be described as follows. All the trays were cleaned and dried before the sieve analysis. A sample of 200 g of material was loaded on the top sieve. The sieve tray assembly was then placed on the shaker table and secured with the shaker tables clamp. The shaker table was set to power 8, and run for 15 min. After sieving, the mass of the powder on individual trays was weighed. A bar chart (Fig. 6(a)) of mass fraction vs. size bins was thus created. Particle diameters (d10, d50, or d90) were read from the cumulative PSD graph using linear interpolation. 5. Experimental conditions The experimental conditions examined for the Comil and hammer mill are presented in Tables 1 and 2 respectively. The range of impeller speeds used in both mills was similar. The milling chamber of hammer mill is larger than that of the Comil, which makes the tip speeds in the hammer mill greater than those in the Comil for equivalent impeller RPMs. 6. Results 6.1. Effect of operational parameters on the flow rate and impeller passes in the Comil
Fig. 1. (a) Conical mill (Comil—Quadro Model # 197); (b) milling chamber with conical round impeller.
The PSD of the milled material is dependent on its residence time in the mill. The residence time of the material in the mill was estimated by measuring the material hold-up in the mill and dividing it by the mass flow rate. Out of the two possible feeding modes (chocked feeding and starved feeding), the Comil was operated under chocked feeding condition. Under the starved feeding condition, flow rate of the incoming material is constant, and hold-up (mass) in mill varies as a function of the operational parameters (screen, impeller speed); whereas under the chocked feeding
86
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 2. (a) Comil screen designs, round holes (R) vs. round-grated holes (RG); (b) impeller design.
condition, the hold-up of the material in the mill is nearly constant, and flow rate is dictated by the operational parameters. The choice between ‘starved’ and ‘choked’ operation is often dictated by process constraints and desired product attributes. For continuous processes, operating at specified mass-rates, milling operations invariably occur under starved conditions, unless, the mass-rate overwhelms the mill. The mill parameters are tuned to achieve the desired product properties. Operating under choked condition enables one to exert more control on the total energy input to the material. Operations in the choked mode are also typified by reduced dusting and fewer fines. The effect of operational parameters on flow rate is shown in Fig. 7(a). Some variations in the hold-up are expected because the PSD of the product in the milling chamber varies as a function of the operational parameters of the mill. These variations in this case are relatively small since the volume of the milling chamber is smaller than the total empty volume in the mill, which includes the milling chamber as well as the feeding section. The residence time (hold-up/flow rate) of the material in the mill was determined by measuring the hold-up and flow rate under steady state. As depicted in Fig. 7(a), the flow rate was found to increase linearly with an increase in the impeller speed. The screen hole size and the screen geometry significantly affected the flow rate. As anticipated, screens with bigger hole-diameters showed higher flow rates. A screen with a combination of round and grated type holes exhibited higher flow rate than the round holed screen with equivalent diameter. Residence time measurement can be used to estimate the total number of blade passes. The number of blade passes applied on the material
during its residence time in the mill was calculated using the relationship given in Eq. (1). Number of impeller passes ðRevsÞ Hold‐up ðkgÞ ¼ Impeller Speed ðRevs=sÞ Flow rate ðkg=sÞ
ð1Þ
The behavior of the number of impeller passes as a function of operational parameters is shown in Fig. 7(b). As expected, the number of impeller passes exhibited the following order—991 μm R N 1575 μm R N 1905 μm R N 1575 μm RG. However, and less expected, the relationship between the impeller speed and impeller passes was strongly affected by the screen size/type. The number of impeller passes was the maximum at the intermediate impeller speed for 991 μm R screen, whereas it decreased with increase in impeller speed for 1575 μm RG screen. Other screens, 1575 μm R and 1905 μm R, showed a relatively constant number of impeller passes, and a slight decreasing trend as a function of impeller speed respectively. These results show that the relationships between impeller speed and impeller passes are non-monotonic, and they need to be characterized individually. With an increase in impeller speed, the impact force exerted by the impeller while it passes through the material bed increases, which essentially decreases the particle size. With increase in the number of passes, the total shear applied on the material increase which also decreases particle size. These observations can be used to interpret the size reduction behavior observed in the mill, and draw conclusions about the underlying milling mechanisms.
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
87
Fig. 3. Hammer mill (Fitzpatrick, Model—DASO6). (a) Hammer mill equipment, (b) position of the impeller and screen, (c) screen design and (d) hammers.
6.2. Effect of operational parameters on PSD The effect of operational parameters on the PSD was analyzed by tracking changes in certain statistical PSD parameters (d10, d50 and d90). A comparison between the two mills (Comil and hammer mill) is shown in Fig. 8.
6.2.1. Effect of screen size In both mills, with an increase in the screen diameter, particle size at the mill discharge increases. For the Comil, as shown previously, an increase in the screen diameter decreases the residence time in the mill, meaning fewer number of impeller passes (breaking action) on the material. A correlation between impeller passes and particle size is shown in Fig. 9. For screens with the same hole shape (R), an increase in the number of impeller passes leads to decrease in particle size. The trend is more visible for particles of greater size. The monotonic relationship between the number of impeller passes and particle size does not hold for screens with different hole shapes. As shown in Fig. 9, for RG type of screens, although fewer number of impeller passes are exerted, an equivalent increase in particle size is not observed.
Qualitatively, the effect of screen diameter in the hammer mill is similar to that of the Comil; however the effect of residence time could not be tested. Residence time measurements for the hammer mill were not available because of the difficulty in measuring hold-up and flow rate under controlled conditions. The effectiveness of the screen was analyzed by calculating α values as a function of the operational parameters (Fig. 10). The definition of α is given in Eq. (2).
α¼
d95 Screen Aperture
ð2Þ
In the Comil, the RG screen provides greater α values than R screens. The RG screen, because of the inherent shape, provides less resistance to the material flow, which increases particle size. In the case of the R screens, the differences between the α values were minor; the 991 μm screen showed slightly greater values. Further comparison between screens is difficult using this data since the screens are also different with respect to the % open area and thickness, and not just the aperture. α values were found to be higher for the hammer mill than the Comil.
88
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 4. Material (a) before and (b) after milling.
6.2.2. Effect of the mill and the screen size on the relative effect of impeller speed (REIS) The effect of impeller speed in the two mills was clarified by analyzing the sensitivity of the particle size distribution to impeller speed, defined as the rate of change of statistical parameters (d10, d25, d50, d75 and d90) as a function of the impeller speed. A linear regression analysis was performed on a dataset of particle size vs. speed at each size cut. The quantity plotted on the Y-axis in Fig. 11 is the slope of this linear fit, normalized by the average particle size. This new quantity was defined as the relative effect of impeller speed (REIS). The equation below was used to compute the REIS values for various screens. Change in dx Change in impeller speed REIS ¼ d50
ð3Þ
The (REIS) value was plotted against the size cut for each screen. The quantity captures the sensitivity of a particle size distribution parameter with changing impeller speed. The larger the modulus of the value, the greater is the impact of the impeller speed. The value assists in making judicious design choices for choosing impeller speeds when constrained by mill capacities and desired particle size distributions. As shown in Fig. 11(a), for the Comil, the relative effect of impeller speed decreases (less negative slope) with increase in size cut. This trend is observed between d25 and d90. For the Comil, a comparison between the screens was made using the REIS values. For the round holed screens, an increase in screen diameter from 991 to 1575 microns increases the REIS. A further increase in the screen diameter decreases the REIS. For the RG type screen, even lower rates were observed. Thus the REIS seems to be a function of
Fig. 5. Sieve shaker.
both the residence time in the mill, and the inherent size reduction rate which is a function of the initial particle size. Typically, the size reduction rate decreases with decrease in particle size, and the residence time in the mill increases with decrease in the screen diameter (except for different screen geometry)—991R N 1575R N 1905R N 1575G. With these two phenomena coupled together, the REIS increases only until the screen diameter reaches 1575 microns; any further increase in screen diameter (decrease in residence time) makes the particles pass through the screen without any significant breaking action. A completely different behavior was observed for the hammer mill. The REIS was found to decrease with increase in particle size and then increase again. The absolute REIS values were much smaller than the Comil for equivalent screen sizes. This behavior suggests that in the hammer mill, the larger particles exit through the screen at a relatively faster rate. Air entrainment in the mill might be playing a role in this case. The hammer mill was operated under the starved feeding condition, as opposed to the chocked feeding condition in the case of the Comil. Once the extrudate material enters the mill, it is broken down into several fragments. These fragments settle on the screen, and are impacted by the hammers at every revolution. While the hammer mill is operating, significant amount of air is entrained, which causes the number of contacts between hammer and particles or particle–particle to decrease. Further size reduction of the initially created fragments happens primarily though abrasion. During this process, once the particles are reduced to a size sufficiently smaller than the screen size, they make an
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
89
Table 2 Experimental conditions for Hammer mill. Agitation rates (RPM)
1419.17 2240.83 3062.5
Screen type (diameter in μm) 1000 R
1500 R
2000 R
X X X
X X X
X X X
Tip speeds (cm/s)
1462.75 2309.64 3156.53
correspond to different screens. The solid markers represent data points corresponding to the Comil, and the open makers represent the data points corresponding to the hammer mill. As shown in Fig. 12(a), for the case of 991 μm screen, d10 does not show any particular trend with respect to d50. However for other screens an increase in d10 was observed as a function of increase in d50. When all the data points corresponding to different screens and speeds are considered together, a plateau in the d10 as a function of d50 was observed. For the case of hammer mill, for individual screens, a clear trend between d10 and d50 was not observed. However, when all the data points are considered together, a relatively linear relationship with a relatively high degree of correlation (R2 = 0.83) was observed. The hammer mill shows lower d10 values (i.e., a broader PSD) than the Comil at equivalent d50. The behavior of d90 as a function of d50 is shown in Fig. 12(b). For the case of the Comil, two linear regions can be identified. The data-points for the 991 μm screen lie on a separate linear curve than
Fig. 6. Calculation of PSD parameters using mass fractions collected on sieves. (a) Raw data of mass fraction vs. size bins; (b) calculation of PSD parameters by linear interpolation.
exit (leak) through the screen. This phenomenon creates a significant amount of fines. This could be a possible milling mechanism in the hammer mill. In the Comil, the number of contacts (both particle–particle and particle–hammer) per unit volume and per time is significantly greater than the hammer mill. This makes the size reduction process more continuous (gradual) as opposed to the hammer mill. The leakage of bigger particles through the screen, which was observed for the hammer mill, is less likely for the Comil, since particles have to pass through the material bed in the mill, and also through the shear zone between impeller and screen.
6.3. Correlations between the statistical particle size parameters The PSD data from the milling experiments can be used for design purposes. In this analysis, in Fig. 12, the d10, d25, d75 and d90 values are plotted against the d50 of the particle size distribution from the individual milling experiments. The d50 of the particle size distribution is used as the design variable. The different markers in the figure
Table 1 Experimental conditions for Comil (‘R’—Round holes, ‘RG’—combination of round and grated holes). Agitation rates (RPM)
1419.17 2281.8 3144.41
Screen type (diameter in μm)
Tip speeds (cm/s)
991 R
1575 R
1575 RG
1905 R
X X X
X X X
X X X
X X X
802.15 1289.77 1777.30
Fig. 7. (a) Effect of Comil impeller speed and screen size on the flow rate; (b) effect of Comil screen size and impeller speed on impeller passes.
90
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 8. Effect of impeller tip speed on the PSD. (a) d10 vs. tip speed, (b) d50 vs. tip speed and (c) d90 vs. tip speed for both mills. Different level curves represent different screens.
the rest of the two screens. However for the hammer mill, a single linear relationship (R2 = 0.99) was observed when all the data points are considered together. The hammer mill shows higher d90 than that of the Comil for equivalent d50 values. Other statistical parameters including d75 and d25 exhibited fairly good linear correlations with d50 for both the mills (Comil: d25—R2 = 0.92, d75 R2 = 0.97, and Hammer mill: d25—R2 = 0.99, d75 R2 = 0.97). Hammer mill shows lower d25 values, and higher d75 values than those for the Comil, again indicating a broader PSD. These results show that milling in a hammer mill results into larger coarse particles (d75 and d90), and smaller fine particles (d10 and d25). For the case of the Comil, the opposite behavior was observed. It produces larger fine particles (d10 and d25), and smaller coarse particles (d75 and d90). The Comil thus produces a narrower PSD. Using this methodology, the desired PSD is selected by comparing one statistical parameter against every other parameter. Once the desired PSD is decided, operating conditions can further be optimized based on the throughput requirements or energy constraints.
the PSD, and normalizing it by the median (d50). For the hammer mill (Fig. 13(b)), the PDI was nearly constant and it did not vary significantly with d50. In the case of the Comil (Fig. 13(a)) the PDI showed three distinct linear relationships depending on the screen size/type used. In the first relationship, the data points corresponding to 1575 μm and 1905 μm round holed screens lie on the same linear curve. The second curve and the third curve correspond to the 1575 μm RG screen and the 991 μm R screen respectively. In all the three curves an increase in the impeller speeds leads to decrease in the PDI. These results show that Comil produces narrower particle size distributions when the desired particle size is large. Both mills produce nearly the same particle size distribution when the desired particle size is smaller. It seems that the PSDs are not scalable as a function of either impeller speed, or screen size. A combination of both the parameters needs to be used in order to scale the PSDs. For the Comil, for screens with similar geometry, if a higher particle size is desired, larger screen and smaller speed are desired. For the hammer mill, the effect d50 on the PDI is different for different screens. It is difficult to deduce any scaling rule using this dataset for the hammer mill.
6.4. Scaling behavior of PSDs 7. Conclusions The scaling behavior of PSDs as a function of operational parameters was analyzed by using the poly-dispersity index (PDI). The PDI was defined by taking the ratio of the difference between the d90 and d10 of
In this study, size reduction behavior of two types of continuous mills, a conical mill—Quadro-Comil, and a hammer mill—Fitzpatrick,
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
91
Fig. 9. Correlation between impeller passes particle size. (a) d10 vs. impeller passes, (b) d50 vs. impeller passes and (c) d90 vs, impeller passes for the Comil.
was characterized for alumina-magnesia extrudates. The effects of operational variables (screen size, shape and impeller speed) on the PSD of milled product have been summarized below. Although, the effect of
Fig. 10. Effectiveness of screen for the Comil and hammer mill.
mill parameters on product properties is a strong function of the material attributes of the feed, the study sheds light on the breakage mechanism in the two mills and consequently, the subtle differences in the effect of milling parameters on process performance between the two mills. Qualitatively the effect of screen size was similar in both mills; an increase in screen size leads to increase in particle size. In the Comil, for screens with identical shape of the aperture, an increase in the number of impeller passes leads to a decrease in particle size. The grated type screens exhibited higher milling capacity than round screens with equivalent apertures. The effect of speed was different in the two mills. For the Comil, except for fines (d10), generally, an increase in impeller speed lead to a decrease in particle size. Smaller particles showed higher sensitivity to impeller speed. For the hammer mill, the nonmonotonic effects of speed were observed across different size cuts in the PSD. Overall the sensitivity of impeller speed to size reduction was less in the hammer mill, and the extrema of the PSD parameters showed higher sensitivity to impeller speed. A comparison between the statistical PSD parameters was performed for process design purposes. If the desired particle size is greater, the Comil shows narrower particle size distributions than the hammer mill; whereas if the desired particle size is smaller, both mills show similar poly-dispersity. In the hammer mill greater amounts of fines are created primarily because of the abrasion of the large particles. Large particles tend to leak through the screen, often without sufficient breaking action. In the Comil, a more gradual size reduction process is
92
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
Fig. 11. Sensitivity of particle size to impeller speed in (a) Comil and (b) Hammer mill.
Fig. 12. Correlations between the statistical parameters of PSD. (a) d10 vs. d50, (b) d90 vs. d50, (c) d25 vs. d50 and (d) d75 vs. d50.
A. Vanarase et al. / Powder Technology 278 (2015) 84–93
93
observed. The creation of fines is primarily through the impact of hammers. The leakage of large particles through the screen is relatively less since the swept area by the impeller against the screen in the Comil is greater than the hammer mill. This study provides design rules for the milling process, under capacity and constraints of PSD parameters. Also a methodology to select optimal combination of operational variables, for a given particle size is presented. References [1] Y. Chen, Y. Ding, D.G. Papadopoulos, M. Ghadiri, Energy-based analysis of milling αlactose monohydrate, J. Pharm. Sci. 93 (2004) 886–895. [2] L.R. Schenck, R.V. Plank, Impact milling of pharmaceutical agglomerates in the wet and dry states, Int. J. Pharm. 348 (2008) 18–26. [3] J.J.A.M. Verheezen, K. van der Voort Maarschalk, F. Faassen, H. Vromans, Milling of agglomerates in an impact mill, Int. J. Pharm. 278 (2004) 165–172. [4] F. Shi, T. Kojovic, J.S. Esterle, D. David, An energy-based model for swing hammer mills, Int. J. Miner. Process. 71 (2003) 147–166. [5] G.K. Reynolds, Modelling of pharmaceutical granule size reduction in a conical screen mill, Chem. Eng. J. 164 (2010) 383–392. [6] L.G. Austin, R.R. Klimpel, P.T. Luckie, Process Engineering of Size Reduction: Ball Milling, Society of Mining Engineers of the American Institute of Mining, Metallurgical, and Petroleum Engineering, Inc., New York, 1984. 22–28. [7] L.G. Austin, A preliminary simulation model for fine grinding in high speed hammer mills, Powder Technol. 143–144 (2004) 240–252. [8] H. Berthiaux, C. Chiron, J. Dodds, Modelling fine grinding in a fluidized bed opposed jet mill: Part II: Continuous grinding, Powder Technol. 106 (1999) 88. [9] H. Berthiaux, J. Dodds, Modelling fine grinding in a fluidized bed opposed jet mill: Part I: Batch grinding kinetics, Powder Technol. 106 (1999) 78.
Fig. 13. Effect of operational parameters on the scaling behavior of PSDs. (a) Polydispersity index vs. median (d50) for the Comil; (b) poly-dispersity index vs. median (d50) for hammer mill.