Axial inlet cyclone for mineral processing applications

Axial inlet cyclone for mineral processing applications

Minerals Engineering 16 (2003) 1375–1381 This article is also available online at: www.elsevier.com/locate/mineng Axial inlet cyclone for mineral pro...

513KB Sizes 0 Downloads 50 Views

Minerals Engineering 16 (2003) 1375–1381 This article is also available online at: www.elsevier.com/locate/mineng

Axial inlet cyclone for mineral processing applications T. Yalcin *, E. Kaukolin, A. Byers School of Engineering, Laurentian University, Sudbury, Ont. Canada P3E 2C6 Received 14 January 2003; accepted 6 September 2003

Abstract An axial inlet cyclone was developed and tested for wet size classification of particulate materials with the purpose of providing an alternative to the tangential inlet cyclone used traditionally in the mineral processing industry. The proposed device shared many of the fundamental features of a conventional cyclone, but had a circular opening on the roof of the cylindrical section and surrounding the vortex finder for feed entry, instead of a tangential inlet. The feed material spiraled down through that opening and generated the necessary spin for separation inside the cyclone. Experimental work was carried out on a sample of copper–nickel mill tailings having a particle size of 91% passing 300 lm, and the performance of the cyclone was evaluated at different inlet pressures, feed pulp densities and vortex finder lengths. In comparison to the tangential inlet cyclone, the axial inlet cyclone provided higher throughputs, coarser cut-sizes in relatively dilute pulps, and greater flexibility and control over the cyclone separation process. Ó 2003 Elsevier Ltd. All rights reserved. Keywords: Mineral processing; Hydrocyclones; Classification; Sizing; Particle size

1. Introduction Many industries enjoy the benefits of hydrocyclone in a diverse range of applications involving solid–solid, solid–liquid, and even liquid–liquid separations (Trawinski, 1976; Svarovsky, 1985). Despite its apparent simplicity, however, the hydrocyclone is a multivariable device and has been the subject of intensive research particularly in the last 50 years or so. Published literature in this area is overwhelming and has to some extent been reviewed by Bradley (1965) and Svarovsky (1984). Various aspects of the hydrocyclone have been thoroughly studied by many researchers in an effort to improve or more efficiently control its performance. Some of the work has been related to the influence of design and operating parameters of the hydrocyclone, some to mathematical modeling and simulation, and others to design modifications. The mineral processing industry makes extensive use of the hydrocyclone particularly in closed grinding circuits, where size classification of the mill discharge is required to remove the sufficiently ground material as * Corresponding author. Tel.: +1-705-675-1151x2252; fax: +1-705675-4862. E-mail address: [email protected] (T. Yalcin).

0892-6875/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.mineng.2003.09.003

finished product, and return the remaining material back to the mill for further size reduction. Choice of a tangential inlet type hydrocyclone has been the standard practice in these applications, but recently certain concerns have been expressed over the throughput of these cyclones and their poor performance in handling concentrates (containing frother) in regrinding circuits. It was to address the throughput issue in particular that the present work on axial inlet cyclone was undertaken, since this type of cyclone is known to deliver high throughputs in other applications, particularly those involving the treatment of gaseous effluents. The axial inlet cyclone has been in use for a long time in industry for the collection of particulates from waste gas streams (Parker, 1977; Ross, 1972; Straus, 1975; Theodore and Buonicore, 1988). It operates on the same principle as the tangential inlet cyclone, i.e. separation is achieved in a centrifugal force field generated by spinning the material inside the cyclone, but the two methods differ in the manner in which they achieve the spinning action. In the tangential inlet cyclone, the material is fed through a tangential slot on the side and near the top of the cylindrical section, whereas in the axial inlet cyclone, the material enters the cylindrical section centrally through the top and around the vortex finder, usually via curved vanes.

1376

T. Yalcin et al. / Minerals Engineering 16 (2003) 1375–1381

2. Experimental 25.4 mm

2.1. Apparatus

101.6 mm

For the experimental work, a hydrocyclone setup was designed and constructed as illustrated in Fig. 1. The setup allowed for the cyclone to be used interchangeably both as an axial inlet cyclone (Fig. 2a) and a conventional tangential inlet cyclone (Fig. 2b), so that comparisons could be made between them on similar bases. The only difference between the two designs was a volume increase of less than 5% in the axial inlet cyclone due to the addition of an axial inlet section, but this was not considered to be significant enough to create any bias in the comparisons. Essentially, the hydrocyclone consisted of a conical section, a cylindrical section, and a vortex finder/overflow assembly. The conical section was fabricated from moulded fiberglass, and had a 15° included angle, which falls mid-way within the typical range of 10–20° angles employed in classifying cyclones (Arterburn, 2000). The cone was 88.9 mm (3.5 in.) wide at the top, and tapered down to a 12.7 mm (0.5 in.) diameter opening at its lower end to form the spigot. The cylindrical section comprised three segments joined together. The lower segment, measuring 25.4 mm (1 in.) in length, was permanently affixed to the top of the cone, while the other two segments were detachable. The middle segment was of variable length, allowing cylindrical sections of different length to be constructed. The upper segment, measuring 76.2 mm (3 in.) in length, incorporated an optional tangential inlet and was capped with an interchangeable plate that had a central opening to accommodate either the vortex finder only or both the vortex finder and the axial inlet, depending on the mode of

50.8 mm 50.8 mm

AXIAL INLET

152.4 mm

TANGENTIAL INLET

88.9 mm 299.7 mm 25.4 mm

15 degrees

12.7 mm

965.2 mm

Bypass

SUMP

PUMP

Fig. 1. Schematic of the experimental hydrocyclone test unit.

operation. All three segments had inner diameters of 88.9 mm (3.5 in.), and were made of clear acrylic tubing. With a middle segment of 50.8 mm (2 in.) length, the total cyclone length (i.e. the cylindrical section plus the

Fig. 2. Views of the experimental hydrocyclone.

T. Yalcin et al. / Minerals Engineering 16 (2003) 1375–1381

conical section) measured 452.1 mm (17.8 in.), which corresponded to a length/diameter ratio of 5, matching Rietema’s (1961) recommendation. The tangential inlet was made from a piece of thick plastic tubing, and fused onto a tangentially carved opening on the side of the cylindrical section, approximately 12.7 mm (0.5 in.) from the top of the section. This inlet provided an immediate tangential path of entry for the feed as it flowed into the cyclone. The inlet had an equivalent diameter of 24.7 mm by area, thus providing an inlet diameter-to-cylinder diameter ratio of 0.28, in conformity with the Rietema (1961) cyclone design. When the cyclone was operated in the axial inlet mode, the tangential inlet was plugged off, and the capping plate of the cylindrical section was exchanged for the one capable of permitting axial entry. The axial inlet consisted of a 50.8 mm (2 in.) dia. acrylic tubing fitted around the 25.4 mm (1 in.) dia. vortex finder. Feed was introduced through a tangential slot near the top of the axial inlet tubing and spiraled down into the cylindrical section (Fig. 3). Material continued spinning after entering the main body of the cyclone. The use of curved vanes to generate the spinning motion, as employed in waste gas treatment applications, was not considered here in order to ensure simplicity of design. Axial entry has the effect of pushing down the feed material, which helps prevent re-entrainment of fines into the incoming feed as they circulate around the vortex finder. The axial inlet design used in this study also allows for some coarse-fine separation to already take place before the feed enters the cyclone. The overflow system consisted of a 25.4 mm (1 in.) dia. vortex finder tubing connected to an inverted ushaped piping of similar diameter. The vortex finder

Axial Inlet Capping Plate

Feed Tangential Slot

Axial Inlet Tubing

Cyclone Capping Plate

Cyclone Body

Fig. 3. Details of the axial inlet.

length, defined as the length of vortex finder tubing that extends into the cylindrical section of the cyclone, could readily be varied by raising or lowering that tubing. The underflow and overflow products were discharged into a sump and then pumped back to the cyclone, thus forming a closed-circuit operation. Cyclone inlet pressure was measured by means of a glycerin-filled mechanical pressure gauge mounted close to the inlet, and was regulated by varying the feed rate via adjustment of valves on the cyclone feed and bypass streams. Since the overflow discharge point was very close to the cyclone and was at near atmospheric pressure, the inlet pressure readings were considered to approximately represent the pressure drop across the cyclone. In order to ensure that the samples taken from the underflow and overflow streams were simultaneous and representative, a duplex rail system was constructed and mounted on top of the sump. The bottom rail was clamped to the sump and remained stationary, while the top rail could be easily slid along the bottom rail in a brisk, smooth motion. Containers were placed upon the top rail and aligned to the flow streams. To facilitate the collection of samples from the discharge streams, suitable lengths of acrylic piping were fitted around the discharge openings, in such a manner as not to disturb the patterns of flow. The total volume of pulp circulating in the system amounted to 15 L. Volumetric throughputs of 106 L/ min in axial inlet mode and 78 L/min in tangential inlet mode were achieved when operating at an inlet pressure of 86.2 kPa gauge (12.5 psig) and a feed pulp density (weight percentage of solids in the solid–liquid mixture fed into the cyclone) of 20%. These throughputs corresponded to solids handling capacities of 1.46 and 1.08 t/ h, respectively. 2.2. Material

Overflow Vortex Finder Tubing

1377

The cyclone feed material used in this study consisted of mill tailings obtained from Falconbridge Limited’s Strathcona Mill, which is located in Onaping, approximately 450 km north of Toronto, in the province of Ontario, Canada. With a capacity of approximately 3 million tonnes of ore per year, the Strathcona Mill processes ore from four of Falconbridge’s Sudbury mines, producing two concentrate products; a copper– nickel concentrate and a copper concentrate, which are subsequently smelted at Falconbridge’s smelters. The Strathcona Mill produces approximately 2.3 million tonnes of tailings per year, 60% of which is used for mine backfill, with the remainder sent to the tailings impoundment area. Hydrocyclones are used for backfill preparation from the tailings. For laboratory tests, representative samples were taken from these tailings. Size analysis of the samples showed a particle size of 91% passing 300 lm sieve size,

Cumulative Wt. % Passing

1378

T. Yalcin et al. / Minerals Engineering 16 (2003) 1375–1381

finder length of 50.8 mm (2 in.). Not all tests were duplicated, but those that were, indicated good reproducibility.

100 90 80 70 60 50 40 30 20 10 0

3. Results and discussion 3.1. Cyclone throughput 10

100

1000

Particle Size (µm)

Fig. 4. Particle size distribution of the cyclone feed material.

with the full size distribution, obtained by a combination of sieving and laser analysis, given in Fig. 4. The specific gravity of the material was determined to be 2.83.

The throughput (capacity) of a cyclone, usually expressed as the volumetric flow rate of cyclone feed, is governed primarily by its size, but a given cyclone can yield a range of throughputs depending on the actual operating conditions employed. Tests conducted with water only revealed a strong influence of inlet pressure (Pi ) on throughput (Q), as illustrated in Fig. 5, which depicts a linear relationship between them, i.e. Q / Pi . Salama and Kizior (1998) derived the following model:

2.3. Methodology

Q ¼ Kc D2:12 Dp0:498 c

Timed representative samples were taken simultaneously from the underflow and overflow streams in order to collect the necessary data for performance evaluations. The basic recorded data included solids and water flow rates, and particle size distributions. Size analyses were obtained by wet-and-dry sieving and laser particle size analysis. Data for the cyclone feed was derived by mathematically combining the underflow and overflow products. Solids flow rate and particle size distribution data provided the necessary information for the construction of cyclone classification curves (also known as efficiency curves, partition curves, Tromp curves, performance curves or selectivity curves), which are plots of solids recovery to underflow, expressed as either fraction or percentage, versus particle size. Tests were conducted first with water only in order to make a rapid assessment of the two cyclones and find out whether a sufficient contrast existed between their performances to justify a further and detailed investigation. Experiments with pulps ensued, using the material described above. Various performance criteria were evaluated, including throughput, cut-size, sharpness of separation, solids and water splits, underflow and overflow pulp densities, etc., but only the first three have been reported here, since these were considered to provide a sufficient basis for comparison between the two cyclone designs. These parameters were assessed as a function of cyclone inlet pressure, feed pulp density and vortex finder length. Inlet pressure was varied between 24.1 kPa gauge (3.5 psig) and 103.4 kPa gauge (15 psig), feed pulp density between 15% and 25%, and vortex finder length between 0 and 127 mm (5 in.). While studying one variable, others were kept constant at their standard values, which were an inlet pressure of 69 kPa gauge (10 psig), a feed pulp density of 20%, and a vortex

where Q is cyclone throughput in US gal/min, Kc is a coefficient dependent on cyclone size, Dc is cyclone diameter in inches, and Dp is pressure drop across the cyclone in psi. Using a Kc value of 0.408, Dc value of 3.5 in., and Pi for Dp, this model yielded an excellent agreement with our data for the tangential inlet cyclone, incurring an average error of only 1.1%. Fig. 5 shows that, at a given inlet pressure, the axial inlet cyclone provided 60–80% higher throughput compared to the tangential inlet cyclone. While the throughput increased with increasing inlet pressure, the recovery to underflow decreased, as seen in Fig. 6. This effect may be explained by the fact that as the inlet pressure is increased, the air core diameter increases (Barrientos et al., 1993; Williams et al., 1995), resulting in an increase in the effective overflow area-to-underflow area ratio. Fig. 7 provides a throughput comparison between the two cyclones on the basis of equal underflow recoveries, and indicates up to three times higher throughput for the axial inlet cyclone.

Throughput (L/min)

120 Axial

100

Tangential

80 60 40 20 0 24.1

34.5

51.7

69.0

86.2

103.4

Cyclone Inlet Pressure (kPa gauge) Fig. 5. Effect of inlet pressure on cyclone throughput (water only).

35

90

Tangential

25 20 15 10 5 0 20

40

60

80

100

25

85 80

20 75 70

15

65 10

60

120

15

17

Cyclone Inlet Pressure (kPa gauge)

23

25

Axial (Throughput)

Tangential (Throughput)

Axial (d50 Size)

Tangential (d50 Size)

Fig. 9. Effect of feed pulp density on throughput and cut-size.

140

3.11:1

120

Axial Tangential

100 80

1.85:1

1.70:1

60

Throughput (L/min)

Throughput (L/min)

21

Feed Pulp Density (%)

Fig. 6. Effect of inlet pressure on recovery to underflow (water only).

40 20 0 15

20

25

Recovery to U/F (%)

Fig. 7. Comparison of cyclone throughputs at similar recoveries to underflow (water only).

Similar relationships were observed between throughput and inlet pressure when the cyclone was operated with pulp. As shown in Fig. 8, which was obtained at a feed pulp density of 20%, throughput increased linearly with inlet pressure, and axial inlet cyclone had 30–40% higher throughput. Feed pulp density did not appear to have a significant influence on throughput within the range of pulp densities studied (Fig. 9). Vortex finder length also had a negligible effect (Fig. 10).

90

120

80 100

70 60

80

50 60

40 30

40

20 20

10

0 40

50

60

70

80

0 90

Cyclone Inlet Pressure (kPa gauge) Axial (Throughput)

Tangential (Throughput)

Axial (d50 Size)

Tangential (d50 Size)

Fig. 8. Effect of inlet pressure on throughput and cut-size.

d50 Size ( m)

Throughput (L/min)

19

120

30

100

25

80

20

60

15

40

10

20

5

0 0

20

40 60 80 100 Vortex Finder Length (mm)

120

Axial (Throughput)

Tangential (Throughput)

Axial (d50 Size)

Tangential (d50 Size)

d50 Size ( m)

0

30

d50 Size ( m)

30

30

1379

95

Axial Throughput (L/min)

Wt% of Feed Reporting to U/F

T. Yalcin et al. / Minerals Engineering 16 (2003) 1375–1381

0 140

Fig. 10. Effect of vortex finder length on throughput and cut-size.

3.2. Cut-size (d50 size) A frequently used parameter to characterize the separation process in a cyclone is the so-called ‘‘cut-size’’ (also known as ‘‘d50 size’’ or ‘‘size of separation’’), which refers to the size of particles, of which half is recovered to the underflow and half to the overflow. The d50 size achieved in a particular application will depend on the design of the cyclone and the operating conditions employed. In this study, variation of d50 size was examined as a function of cyclone inlet pressure, feed pulp density, and vortex finder length. As seen in Fig. 8, the inlet pressure had a significant effect on the d50 size in axial inlet cyclone, while having minimal impact in tangential inlet cyclone. The former can be viewed as an advantage, since that would provide more flexibility in cyclone operation and allow for the inlet pressure to be used as an effective means of controlling the separation process. A similar trend was observed with feed pulp density, as illustrated in Fig. 9. With increasing feed pulp density, the d50 size tended to increase more in axial inlet cyclone than it did in tangential inlet cyclone. More importantly, the axial inlet cyclone provided much larger

T. Yalcin et al. / Minerals Engineering 16 (2003) 1375–1381

cut-sizes than the tangential inlet cyclone at any given pulp density. This implies that, in the axial inlet cyclone, coarse size separations are possible with relatively dilute pulps, while a similar separation in the tangential inlet cyclone would require significantly denser pulps, with adverse consequences. Braun and Bohnet (1990) point out that in hydrocyclones ‘‘an increase in feed concentration generally leads to a coarse cut-size, reduced sharpness of separation and a rise in pressure drop.’’ Fig. 10 displays the effect of vortex finder length, and demonstrates a generally decreasing cut-size with increasing length of the vortex finder. This may be due to the longer vortex finder providing more time for particles to be moved away from the vortex. A number of mathematical models are available for the prediction of d50 size, such as those developed by Dahlstrom (Bradley, 1965), Plitt (Plitt, 1976; Flintoff et al., 1987), Lynch and Rao (Lynch, 1977), and Bradley (1960). These have been tested and the Dahlstrom’s model was found to provide the best agreement with our data for the tangential inlet cyclone, giving estimates of d50 size with an average error of 2.7 lm. This model is of the form:  0:5 81ðDo Di Þ0:68 1:73 d50 ¼ rq Q0:53 where d50 is in lm, Do and Di are the overflow and inlet diameters respectively in inches, Q is the feed rate in US gal/min, and r and q are the specific gravities of the solids and the liquid respectively in g/cm3 . None of the above-mentioned models was found applicable to the axial inlet cyclone. 3.3. Sharpness of separation Ideally, separation in a cyclone should be such that all particles larger than the d50 size report to the underflow and all those smaller report to the overflow. In reality, this is never achievable. Both the underflow and overflow products will invariably contain some misplaced material, the extent of which is a measure of the sharpness (efficiency) of separation. Such efficiency is usually displayed in the form of a classification (efficiency) curve mentioned earlier. In these curves, the reading of particle size at 50% recovery represents the d50 size, and the slope of the curve is an indication of efficiency, with larger slopes denoting higher efficiencies. It has been recognized that a part of the material entering the cyclone escapes the classification action and reports to the products unclassified. In order to assess the true classification efficiency, it is necessary to eliminate the effect of such material and obtain a corrected efficiency curve. As suggested by Kelsall (1953), corrections can be made via the expression Rc ¼ ðRa  Rw Þ= ð1  Rw Þ, where Rc is the corrected recovery of solids to

1.0

Fractional Recovery to U/F

1380

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.1

1.0

10.0

d/d50c Axial (Data)

Tangential (Data)

Axial (Fit)

Tangential (Fit)

Fig. 11. Reduced efficiency plot.

underflow at a given particle size, Ra is the corresponding actual recovery, and Rw is the recovery of water to underflow, with all recoveries expressed as fractions. Especially when making efficiency comparisons between different cyclones or conditions of operation, it is often useful to plot the efficiency curve using the d=d50 values on the particle size axis, since all curves would then pass through a common point located at d=d50 ¼ 1 and recovery ¼ 0.5. The resulting curve is known as the reduced efficiency curve (Bradley, 1960). On the basis of above considerations, efficiency comparisons were made between the axial and tangential inlet cyclones tested in this study. Fig. 11 shows a typical reduced efficiency plot, displaying similar performances for the two cyclones. Curve fitting was applied using the efficiency model that was proposed by Reid (1971) and Plitt (1971). The model is of the form Rc ¼ 1  exp½0:6931ðd=d50 Þm , where Rc is the corrected recovery and m is a classification index, representing the slope of the efficiency curve. The m values, determined using Excel’s Solver, were 0.85 for the axial inlet cyclone and 0.78 for the tangential inlet cyclone. Thus, the sharpness of separation appears to be slightly in favour of the axial inlet cyclone. Furthermore, it should be noted that in order to achieve similar cut-sizes with the tangential inlet cyclone as those obtained with the axial inlet cyclone, the former would have to be operated at much higher pulp densities, as is often the case in many mineral grinding circuits, and that would cause deterioration of the sharpness of separation (Braun and Bohnet, 1990) as mentioned earlier.

4. Conclusions This study has demonstrated that the axial inlet cyclone may offer a viable and beneficial alternative to the tangential inlet cyclone for wet size classification in the

T. Yalcin et al. / Minerals Engineering 16 (2003) 1375–1381

mineral processing industry. One of the advantages that can be derived from this type of cyclone is increased throughput, which could result in reduced size or number of cyclones needed for a given application. Within the experimental conditions employed, throughput increases of up to 40% have been realized in comparison to the tangential inlet cyclone. The axial inlet cyclone also makes it possible to achieve separations at coarse cut-sizes even in relatively dilute pulps, while a similar separation in a tangential inlet cyclone would normally require feeds with high solids concentration. Furthermore, the inlet pressure was found to be a critical parameter in the operation of the axial inlet cyclone, having a significant influence on both the throughput and the d50 size. This would suggest that the cyclone performance could be conveniently monitored and controlled through regulation of inlet pressure. It is expected that the development of suitable mathematical models for the axial inlet cyclone would enable the unit to be scaled up to meet industrial throughput requirements. Future work in this area may also involve a comparative study of flow fields in the two cyclones, with the aid of computational fluid dynamics techniques. Such additional studies might help explain the reasons for their differences in throughput and cut-size. Acknowledgements This project was made possible through the financial supports of MMO (Manufacturing and Materials Ontario) and Falconbridge Limited. The assistance of Norman Lotter (Falconbridge) and Mike Grace (MMO) in obtaining these supports is greatly appreciated. The authors also thank Peter Wells of INCO Ltd for his role and encouragements in the initiation of this study. References Arterburn, R.A., 2000. The sizing and selection of hydrocyclones. Available from (last accessed: April 2003). Barrientos, A., Sampaio, R., Concha, F., 1993. Effect of air core on the performance of a hydrocyclone. In: Proceedings––XVIII Interna-

1381

tional Mineral Processing Congress, Sydney 1993, vol. 1. The Australasian Institute of Mining and Metallurgy, Parkville, pp. 267–270. Bradley, D., 1960. Design and performance of cyclone thickeners. In: Proceedings––International Mineral Processing Congress, 1960 (London). The Institution of Mining and Metallurgy, London, pp. 129–144. Bradley, D., 1965. The Hydrocyclone. Pergamon Press, Oxford. Braun, T., Bohnet, M., 1990. Influence of feed solids concentration on the performance of hydrocyclones. Chemical Engineering & Technology 13 (1), 15–20. Flintoff, B.C., Plitt, L.R., Turak, A.A., 1987. Cyclone modelling: a review of present technology. CIM Bulletin 80 (905), 39–50. Kelsall, D.F., 1953. A further study of the hydraulic cyclone. Chemical Engineering Science 2 (6), 254–272. Lynch, A.J., 1977. Mineral Crushing and Grinding Circuits: Their Simulation, Optimization, Design and Control. Elsevier, Amsterdam. Parker, H.W., 1977. Air Pollution. Prentice-Hall, Englewood Cliffs, NJ. Plitt, L.R., 1971. The analysis of solid–solid separations in classifiers. CIM Bulletin 64 (708), 42–47. Plitt, L.R., 1976. A mathematical model of the hydrocyclone classifier. CIM Bulletin 69 (776), 114–123. Rietema, K., 1961. Performance and design of hydrocyclones––IV: Design of hydrocyclones. Chemical Engineering Science 15 (3 and 4), 320–325. Reid, K.J., 1971. Derivation of an equation for classifier-reduced performance curves. Canadian Metallurgical Quarterly 10 (3), 253– 254. Ross, R.D., 1972. Air Pollution and Industry. Van Nostrand Reinhold Company, New York. Salama, A.I.A., Kizior, T., 1998. Hydrocyclone model simulation: a design tool for dewatering oil sands plant tailings. Paper No. 1998.100, presented at the 7th UNITAR Conference on Heavy Crude and Tar Sands, Beijing, 1998. Available from (last accessed: April 2003). Straus, W., 1975. Industrial Gas Cleaning, second ed. Pergamon Press, Oxford. Svarovsky, L., 1984. Hydrocyclones. Technomic Publishing, London. Svarovsky, L., 1985. Solid–Liquid Separation Processes and Technology. Elsevier, Amsterdam. Theodore, L., Buonicore, A.J., 1988. In: Air Pollution Control Equipment, vol. I: Particulates. CRC Press, Boca Raton, FL. Trawinski, H., 1976. Theory, applications, and practical operation of hydrocyclones. Engineering and Mining Journal 177 (9), 115–127. Williams, R.A., Ilyas, O.M., Dyakowski, T., Dickin, F.J., Gutierrez, J.A., Wang, M., Beck, M.S., Shah, C., Rushton, A., 1995. Air core imaging in cyclonic separators: implications for separator design and modeling. The Chemical Engineering Journal and the Biochemical Engineering Journal 56 (3), 135–141.