P e e r R e v i e w e d Separation performance of centrifugal cleaners GEOFF COVEY SUMMARY Although centrifugal cleaners are widely used in the pulp and paper industry, there is very little information available on their relative performance with contaminants of different sizes and densities. This problem has been addressed in the minerals industry for many years, and this paper shows how the methods used there can also be employed when contaminants are present in paper-making pulp suspensions, and how estimates of performance can, if necessary be made without extensive experimental data. INTRODUCTION Centrifugal cleaners have been widely used in the pulp and paper industry for many years as a means of removing small contaminants. Initially they were used only for removing dense material (such as sand and dirt particles) but since the introduction of ‘reverse cleaners’ they have also been used for the removal of lowdensity contaminants – particularly plastic fragments. Common applications include: - In pulp mills to remove sand and grit. - In bleach plants to remove ‘dirt’. - In recycled fibre plants to remove both heavy and light contaminants. - In paper machine stock preparation for final removal of contaminants. In recent years there has also been many papers suggesting the use of centrifugal cleaners to fractionate fibre or to remove fillers from recycled fibre. Comparatively little has been written on the separation performance of hydrocyclones. There is a good deal on the design of systems to give good rejection of contaminants with minimal fibre loss, and there is quite a lot of published information on empirical studies of fibre segregation, but very little on prediction of efficiency of removal of contaminants of various sizes. This information can be very important as contaminants may be present in quite specific size ranges, and although a given arrangement may be effective in removChairman, Covey Consulting Pty. Ltd. 1st Floor, 660 High St. Kew VIC 3102 January 2009 ing sand that is fairly coarse, it may not be effective in removing the same sand after it has been subjected to attrition. The same issue is faced by the minerals industry, which also uses hydrocyclones to separate minerals of different densities and/or sizes. In that industry the quantification of separation can be even more difficult as both mineral species are likely to have fairly wide size distributions. (Note that in the case of pulp fibres, the size range of “good” fibres can be comparatively small.) The minerals industry has successfully developed calculation techniques to quantify the separation, and these techniques can also be applied to centrifugal cleaners. The cross over of the technique from one industry to another is made easier because not only do ‘centrifugal cleaners’ in the paper industry and ‘hydrocyclones’ in the minerals industry work in the same way, but they are often made by the same manufacturers and in many cases the designs of units supplied to the two industries are essentially the same. CENTRIFUGAL CLEANERS Centrifugal cleaners (or hydrocyclones) separate components of the feed by centrifugal action. A schematic of a typical centrifugal cleaner is shown in Figure 1. Unlike a centrifuge, this centrifugal action is not induced by rotating the equipment, but by introducing the feed stream at relatively high velocity, tangentially into a cylindrical body. This creates a vortex that tends to cause high-density components to move to the wall. The lower portion of the cyclone consists of a convergent cone (although this is not theoretically necessary). Material collected at the wall (the high density fraction) is discharged from the bottom of the cone. The bulk of the flow forms an inner vortex that rises to the top of the unit and discharges through a central pipe (the vortex finder). Reference has been made to dense material moving to the wall, and it is always the case that denser components move in this way, and that lighter components of the feed are slower to migrate to the wall and,, being closer to the axis, are discharged through the vortex finder. It is important to realise that when solid components are being separated, separation is not according to differences in density, but to differences in settling velocity in the cyclone. Settling velocity is a function of particle density relative to water, particle size and (to a lesser extent) particle shape. In normal use, all of the components of interest in a centrifugal cleaner are denser than the suspending fluid (water) and all solid components will tend to be collected from the bottom (‘rejects’) outlet. This includes the fibre, and the overall effect is the well-known ‘thickening effect’ whereby the consistency of solids in the rejects stream is typically about twice that in the feed stream. In normal operation therefore, with all of the solids denser than water, all particles are tending towards the wall. If an infinitely long residence time were available, all of the solids would report to the ‘rejects’ outlet and no segregation would occur. In a real system, what is relied on is differences in the settling rates of different solid components. A cross section of the cylindrical section of the hydro-cyclone is considered in Figure 2. Accepts Vortex finder Tangential inlet Rejects Fig. 1 Schematic of a typical centrifugal cleaner 31 P e e r R e v i e w e d The first two of these factors result in more contaminants appearing in the accepts than would otherwise be the case (and hence rejection of less contaminants) and the last factor results in the rejection of more good fibre than would otherwise occur. REVIEW OF LITERATURE Contaminants denser than water move towards wall Fig. 2 Cross-section of a centrifugal cleaner showing swirling flow and separation of dense material near to the wall. Obviously the particles to be removed must migrate through a finite distance of rotating fluid to the wall, approximating the width of the inlet flow channel.(or somewhat less it they enter the cyclone at a point closer to the wall). Ideally, the residence time of the hydrocyclone can be selected so that there is time for the fast settling particles (contaminants) to reach the wall and be discharged as rejects, while none of the slower settling particles (good fibre) have time to reach it. In practice, of course, some of the fibres enter the hydro-cyclone close to the wall and are rejected, and some of the contaminants enter the cyclone far from the wall and do not have time to migrate to it. Therefore separation is not perfect. Further factors that also contribute to imperfect separation are: • Short-circuiting – some material short-circuits from the rotating outer region directly to the inner core and out through the vortex finder. Little or no separation occurs with this component. • Turbulence – varied factors can cause the formation of large eddies which sweep fluid from close to the wall back into the bulk region, and so negates the separation already performed on this fluid. • Fluid discharge with rejects – It is necessary to discharge some of the inlet liquid with the rejects just to maintain movement of the rejects and avoid plugging of the bottom outlet. This fluid will have a composition similar to that of the accepts fluid, and so it represents an inevitable loss of separation efficiency. 32 Appita Journal There is a very large body of literature on hydrocylones in general and their use as cleaners in the pulp and paper industry. Therefore the following review is not intended to be exhaustive, but merely to give examples of the type of work undertaken, with more detail on their use for the removal of foreign bodies, which is the main subject of this paper. The first description of a hydrocyclone is apparently in the 1891 patent of Bretney, which decribed its use for removing impurities from water (1). The first proposal for use of the device in the paper industry was to remove high density contaminants form stock was by MacNaughton in 1906 (2). Actual use in the industry started in the 1930’s, but was not widespread until the 1950’s. Publications on hydrocyclones and cyclones (1) have tended to be of two types, focussing on either the flow patterns inside the units or on their separation performance. Work on separation performance generally falls into three categories: the removal of shives and other ‘dirt’; fibre fractionation; and the removal of foreign material such as sand and grit. Flow studies Studies (3) have shown various flow imperfections, such as short-circuiting and circulatory eddies that can limit performance. It is also known that the velocity field in both cyclones and hydrocyclones is not stationary but displays quasiperiodic fluctuations that are generated by the precession of the vortex core (PVC). Zhao and Abrahamson performed a CFD analysis of flow in conical cyclones (4). Their model showed some previously observed features such as short-circuiting and reduction of velocities in the axial direction. However, agreement with physical measurements was often only ‘reasonable’ and because the model was axisymmetrical and steady state, it did not incorporate effects such as PVC. It was also restricted to inviscid flow, which reduces its application to hydrocyclones. Wang et al (5) used a more complex turbulence model than earlier workers and used a model that allowed for asymmetrical flow patterns. This resulted in a more realistic model and indicated that the flow in the vortex in particular was not symmetrical with the physical axis of the unit. Gong and Wang (6) took the approach further by using a unsteady solver, which allowed them to predict some of the effects of PVC. Kegge (7) modelled an axial flow hydrocyclone (the same flow shape as an end-entry cleaner) by CFD, and found that even though there was no enforced asymmetry as in a conventional side-entry hydrocyclone, asymmetrical effects were important and his model was limited in its flexibility. Ko (8) also performed a CFD analysis of an axisymmetrical hydrocyclone for pulp and paper applications, but his analysis did not include fibres in the liquid. Ko’s thesis includes an extensive review of about one hundred papers relating to numerical studies in hydrocyclones published prior to 2005. Shive removal Tomlinson (9) found that when considering the removal of shives and dirt from good fibre, it is the ratio of centrifugal force to angular velocity gradient that is important, and that this permits the use of larger cleaners than had previously been considered necessary for such purposes. By varying the dimensions and proportions of a cleaner he could optimise for the removal of dirt or shives of particular sizes. For large shives, of density similar to that of the pulp fibres, shear effects became very important in determining whether they would be rejected. Hill (10) et al measured shive removal by centrifugal cleaners (and also by screens) and related the cleaning efficiency to the rejects rate, but their findings were specific to a particular pulp and cannot be used for predictive purposes. Corson and Tait (11) presented a series of correlations to predict the outlet consistencies and shive contents of Pinus radiata mechanical pulps. The correlations do not consider differences in removal efficiencies of different sizes of shives or the efficiency of removal of non-fibrous material. Fibre fractionation It was discovered in the 1960’s that hydrocyclones could fractionate pulp to give streams enriched in spring or summerwood (12). Vol 62 No 1 P e e r Paavilainen (13) investigated the separation of summerwood and springwood fibres from Finnish softwood pulp in a hydrocyclone at 0.1% consistency. She found that there was a useful fractionation of the pulp, with the coarser summerwood being concentrated in the rejects and there was also a smaller effect of concentration by length. Wood et al (14) found that for mechanical pulps, fractionation was dependent on specific surface area. However, for unrefined chemical pulp of substantially uniform length and fed at very low consistency (0.05% to minimise flocculation) Li et al (15) found that apparent density of individual fibres was the main factor in controlling fractionation. They found that the results could be described reasonably well by a partition curve. Kanis (16) evaluated the available data and concluded that, when other properties are equal, mechanical pulps were fractionated on the basis of specific surface area, and chemical pulps on the basis of fibre coarseness. Wang et al (17) developed a CFD model of a cleaner in use as a fibre fractionator working at low consistency and compared its predictions with measurements on nylon fibres at 0.3% consistency. The model gave good prediction of separation in by coarseness, but only moderate prediction of separation by fibre length. They attributed the deviations as due to flocculation, which is not considered in their model. With wood fibres at practical operating consistencies, flocculation will be much more significant than in their tests. Park et al (18) studied the separation of beaten and unbeaten softwood kraft fibres at 0.1% consistency. They calculated slip velocities for typical fibres and showed that fractionation could be largely related to differences in slip velocities. This was consistent with their confirmation that as fibre coarseness increased, so did fibre rejection. They also found that rejection increased with fibre length but rejected this finding as an artifice as it was contrary to CFD results and their own theoretical treatment. In fact, if extended their analysis would have shown that increasing fibre length will result in increased rejection, other things being equal. Further, although the CFD analysis that they relied on did not predict an effect from fibre length, the experimental work of these workers did find fibre length to have an effect. There is a considerable body of January 2009 R e v i e w e d research on the various factors that affect fibre fractionation. Further discussion of this is beyond the scope of the present paper, but the literature is reviewed by Bergstrom (19). cles of different sizes to be removed by a hydrocyclone is the empirical expression of Lynch and Rao (23) (see Equation 1, below). The value of this method has been confirmed by a number of other workers. Impurity removal Lepola et al (20) investigated the removal of 10-40µm particles of glass from spruce and birch pulps by a centricleaner by a radioactive activation technique. They measured removal efficiency under various conditions and presented plots for the effect of size, pulp consistency and operating pressure. However, they did not present any fundamental analysis of the system, or any predictive results. Ferguson (21) performed a theoretical mathematical analysis for the removal of impurities. He assumed that the inner vortex behaves as a rotating body, that the outer vortex is a free vortex and a powerlaw viscosity model (although he noted that pulp suspensions are visco-elastic). The model was used to predict the effect of fibres on the flow patterns inside the hydrocyclone and then particle separation efficiency was calculated using this velocity field. The results from the model were of the same general form as is observed in practice, but are not particularly useful for predictive purposes. Fishhook effect. It is well established that as impurity particle size is reduced, the removal efficiency passes through a minimum and then rises again – this is commonly referred to a the ‘fish-hook’. The effect has been studied by various workers and reviewed by Neese et al (22). This effect typically applies to particles smaller than 5 mm. For particles this small, the separation efficiency is low, and the boost due to the fish hook effect does not change this dramatically in most cases. The effect is important in fine grinding of minerals, but is not thought to be of importance in pulp and paper applications. Therefore it will not be considered further here. Partition curve. The most widely used method for predicting the fraction of parti- PARTICLE SETTLING RATES As discussed above, the major factor in separation is the settling rates of each of the particles of interest. For the purposes of this paper, fibres of eucalyptus and pinus radiata, and contaminants of sand and black coal will be considered. Sand is selected as a common dense contaminant (density about 2600 kg/m3) and black coal as a less dense particle (density about 1350 kg/m3, just slightly higher than that of cellulose fibres at about 1100-1200 kg/m3). To calculate the settling velocities it is necessary to know the centrifugal acceleration in the hydrocyclones. This can be calculated on a theoretical basis, but results are not always reliable because of the effect of friction in slowing the liquid flow. According to Gulichsen (29) centrifugal forces may theoretically be about 800 g (i.e. about 800 times the acceleration due to gravity) but are usually somewhat less in practice. For the present purposes a centrifugal acceleration of 500g has been used. Performing calculations at a variety of centrifugal accelerations (and under normal gravity) shows that although the absolute settling velocity changes markedly, the ratio of settling velocities of the various species does not change very much. Therefore, the effect of the spinning action is not to increase separation so much as to accelerate the process by which it occurs. Particles of coal and sand both have dimensions that are of similar magnitude in all directions (isotropic). Therefore, the simple methods for calculating terminal settling velocities that are used for spherical particles can be used (and the results presented below are for ‘equivalent spherical particles’). Using the method in Coulson and Richardson (30) for terminal settling velocities gives the values presented in Table 1. Table 1 Terminal settling velocities of various size contaminants under typical centrifugal cleaner conditions (500g, water at 60°C). Particle size mm Terminal settling velocity coal Terminal settling velocity quartz Ratio of Term. settling velocities coal/quartz m/s m/s 0.05 0.1 0.2 0.5 1 0.22 0.64 0.34 0.46 1.20 0.38 0.85 2.17 0.39 1.60 3.43 0.47 2.27 4.85 0.47 33 P e e r R e v i e w e d Table 2 Terminal settling velocities of wood fibres at various fibre densities (same conditions as Table 1). Fibre density Pine Length Width Equivalent diameter Shape factor Terminal settling velocity kg/m3 1200 1150 1100 mm mm mm 3 0.044 0.41 0.084 0.68 3 0.044 0.41 0.084 0.60 3 0.044 0.41 0.084 0.50 1.1 0.02 0.167 0.094 0.38 1.1 0.02 0.167 0.094 0.34 1.1 0.02 0.167 0.093 0.27 m/s Eucalyptus Length Width Equivalent diameter Shape factor Terminal settling velocity mm mm mm m/s Table 3 Settling velocities of wood fibres of density 1150 kg/m3 and sizes of particles that settle at the same rate. Settling vel m/s Coal µm Sand µm 0.60 0.34 130 75 45 30 Pine Eucalypt Pulp fibres are obviously far from isotropic and it is inappropriate to calculate their terminal settling velocities using the method used for contaminants. There are a number of methods that have been presented for predicting the drag or terminal settling velocity for nonspherical particles (31-33). For the present purposes, the method presented in Coulson and Richardson (30), which is based on the work of Heywood (32) and of Heiss (33) has been used. The method is based on characterising particles as they lie in their most stable position and using a characteristic diameter equal to that of a circle having the same area as the projected area of the particle. As shown in Fig 4 of Heywood’s paper (32), this permits simple modifications to the equations for spherical particles to permit representation of nonspherical particles. The results of Heiss (33) in particular covered isotropic, disc and rod shaped particles. An alternative, more recent approach is that of Haider and Levenspiel (31). This method was based on data for isotropic and disc shaped particles only, and therefore should be used with extreme caution for rod-shaped particles. (Unfortunately their results do not appear to agree with those of other published methods even for spherical particles, so this approach has been neglected for the purposes of this paper). 34 Appita Journal Although the density of cellulose is well known, the density of individual fibres is less definite. This depends on the size of the lumen (a large lumen will reduce the fibre density) and also the degree of fibrillation (a high degree will reduce the effective fibre density). Therefore, in the calculated values in Table 2, settling velocities are presented for typical pine and eucalyptus fibres at densities of 1100-1200 kg/m3. Clearly, separation will only occur when there is a difference in settling velocities. We have calculated the size of particles of coal and of sand that will have the same terminal settling velocities as wood fibres of density 1150 kg/m3. (Table 3) The results in Table 3 show that particles of sand smaller than about 45 µm cannot be separated from pine fibre under the assumed conditions. This also implies that particles smaller than this will actually be more concentrated in the accepts than in the feed (so very fine particles might be removed by means of reverse cleaners). From these simple calculations, it is apparent that very large particles of sand with settling velocities much greater than those of wood fibres can be readily separated, that particles of 30-45 µm will not be separated at all, and that particles of intermediate size will be separated to some extent. However, this simplistic approach is insufficient to determine the degree of separation of intermediate size particles. REDUCED-RECOVERY CURVES The need to determine the efficiency of removal of ‘intermediate’ size particles is regularly faced in the minerals industry where cascades of hydrocyclones are used. In this case the object is to separate minerals initially present at similar concentrations, and sometimes two saleable products will result. The extent to which particles of various sizes can be removed may be estimated by the method of Lynch and Rao(23) which is commonly used in the mineral processing industry and which is described in SME Mineral Processing Handbook (p 3D-22 et seq.) (34). Here the proportion of a component rejected is reported to vary according to the equation: [1] Where: Y is the fraction of a particular size passing to the rejects stream α is a characteristic of the particle-fluid combination and of the cleaner configuration. x is the ratio of the diameter of the particle of interest to the diameter of the particle which passes equally to the accepts and the rejects (usually designated d50). The curve that can be fitted by the equation is a partition curve, and is commonly known as the ‘reduced-recovery curve’. Obviously, this equation is applicable for particles that only differ in size. The density of the particles must be substantially the same, and the shapes must also be similar to the extent that the effect of shape on settling velocity is approximately uniform. The parameter α is determined by fitting available data on rejects and accepts at different sizes for a particular installation. According to the SME Handbook, the value of α is usually in the range 2.5 to 6, and is most commonly 3 to 4. There is limited data available on the size of sand typically removed by cleaners in the paper industry and obtaining such data requires experimental equipment that is not readily available. However results by Kadant (35) give sufficient information to calculate a value for α and to calculate the rejection efficien- Vol 62 No 1 P e e r Table 4 Performance of two commercial cyclones. Cleaner A Cleaner B 10 2 20 80 10 88.89 0.111 8 2.8 22.4 77.6 14.4 84.35 0.157 Rejects % Thickening factor Pulp to rejects % Pulp to accepts % Corrected pulp to rejects % Corrected pulp to accepts % Y cies of coal and sand particles of various sizes. From this data it is found that with a conventional cleaner (as used in most pulp and paper mills) there is about a 9293% rejection of 100µm sand at typical operating consistency of 0.6-0.9% of hardwood pulp. This result can be combined with information on sizes of coal and sand particles having the same settling velocity as pulp fibres and typical cleaner operating parameters to calculate the performance of cleaners in removing coal particles of various sizes. To determine the value of the parameter α, one proceeds by the following steps: 1. Part of the reduced recovery curve can be deduced from the thickening effect with pulp in a known centrifugal cleaner. Data for two commercial cyclones are given in Table 4 As before, Y in Table 4 is the fraction of a particular size passing to the rejects stream. The ‘corrected pulp to rejects’ value is a figure allowing for the fact that some of the fibre in the rejects stream is entrained with the accompanying water, rather than there as a result of classification – so an amount is subtracted from the rejects pulp equivalent to the amount of pulp that would be in the same volume of accepts. (In mineral processing applications, there can be some uncertainty as to whether the concentration applied here should be that of the feed or the accepts, or some intermediate value. However in normal pulp cleaner operation, the consistency of the accepts is not much lower than that of the feed, and it makes little difference which of these values is used.). 2. The value of α is then determined by assuming that it will be the same for both pulp and contaminants in a given cyclone. This assumption is not strictly correct, but it provides a reasonable approximation in the absence of better experimental data. (Essentially it is assumed that the contaminant size having the same settling velocity as the pulp fibres will also be rejected at the same rate as the pulp.) The value of α is then found by fitting the general reduced recovery curve (equation 1) to the two points corresponding to: • the equivalent of the pulp rejects rate for the particular case (as shown in Table 4);and • the rejection of 92-3% of 100µm sand 100 R e v i e w e d (or same rejection rate of another contaminant having the same settling velocity as 100µm sand). Because of the highly non-linear nature of equation 1, it is convenient to substitute one of the points into the equation and then solve the remainder numerically. 3. For the data used here, it was found that the best fit for Cleaner A was with α = 3.0, and for Cleaner B with α = 3.6. Both of these values lie within the most common range quoted by SME as 3-4. Figure 3 presents a graph showing that the data from the two types of cyclone give quite similar results. This is not surprising as they are of similar geometric proportions, and much of the difference relates to set-up for operation. Figure 4 shows the relative performance of a centrifugal cleaner in removing contaminants of sand and coal of various sizes. Clearly, although both particles are more dense than pulp fibres, the performance in separating them is very different because of the differences in the densities of the two contaminants. One stage of cleaning will remove more than 90% of 100 µm sand particles, but for coal an equivalent % removal is only obtained with particles larger than 250µm. This shows that contamination with fine, but relatively low density particles can present difficulties for centrifugal cleaners. It is often not recognised that only the first stage of cleaners in a system removes contaminants. All subsequent stages only work to recover good fibre and so to reduce the losses from the system (in the process they also ‘recover’ some of the contaminants with the good fibre and so reduce the rejection efficiency of the system slightly.) 90 % Reporting to rejects % Reporting to rejects 100 80 60 Cleaner B 40 Cleaner A 20 Average 80 70 60 50 40 Sand 30 Coal 20 10 0 0.000 0 0.050 0.100 0.150 0.200 0.0 Particle size, mm Fig. 3 Reduced –recovery curves for sand using rejects data from two commercial cyclones and α = 3.6. January 2009 0.1 0.2 0.3 0.4 Particle diameter, mm Fig. 4 Reduced recovery curves for sand and coal for α = 3.4 35 P e e r R e v i e w e d The only ways that the cleaners can remove more of a contaminant are by: • Rejecting more pulp in the first stage so that more contaminants are also rejected. This approach requires the use of additional stages of cleaning if fibre loss is not also to increase. Use of a double first stage of cleaning whereby two sets of cleaners (of the same size) are operated in series so that the accepts from one set is re-processed in a second stage. In theory this appears quite attractive (even if expensive) as cleaners work on a statistical basis, and if single screening will remove (say) 90% of a contaminant, then double cleaning will remove 99% in total. Unfortunately it is found that with this type of arrangement, the second stage cleaners are less effective than the first. Although the first cleaner will theoretically remove 90% of a particular size, the 10% it does not remove is often more difficult to treat in the next step. The reduction in apparent performance when either of these approaches is used is significant, but not necessarily sufficiently so as to make double cleaning impractical. The problem is that in the paper industry we do not normally have reduced-recovery curves (or grade-efficiency curves) and there is a tendency to just look at the quantity of contaminant remaining rather than its size distribution. The reality is that the portion of contaminant passing the first step of cleaning is the finest part, and additional passes will not be very effective in removing this. It should also be noted that the treatment given above has been based on the requirement to remove contaminants from uniformly sized fibres. The same approach can also be used to predict the separation of shives, fines or fibres of different species or morphology. The principles described can still be applied, but as the fibres may be of different aspects or effective densities, the actual separation will be on the basis of equivalent diameters (i.e. diameter of the spherical particle of the same density and having the same settling rate) rather than on for example fibre length. CONCLUSIONS Although centrifugal cleaners have been widely used for many years to remove dense contaminants from pulp, there has 36 Appita Journal been little information on the relative removal efficiency of contaminants of different sizes and densities. This paper has explained how a relatively simple technique, which has been used for many years in the minerals industry with the same equipment, can be used to predict the grade efficiency for contaminant removal. Ideally, the parameter α should be determined experimentally for the particular cleaner and contaminant. However in the absence of better information, an estimate of removal efficiency that is adequate for most purposes can be obtained by assuming a value of α of 3-4. REFERENCES (1) Bretney, E.:US Patent US453105 (1891). (2) MacNaughton, J, : US Patent 813984 (1906) (3) Ohtake, T., Usuda, M., Kadoya, T. – A fundamental study of hydrocyclones Part 1. Flow pattern in the hydrocyclone, Japan Tappi J. 41(2), 60 (1987). (4) Zhao, J.Q. and Abrahamson, J. – The flow in Conical Cyclones Proc. 2nd Intl. Conf. on CFD in Minerals and Process Industries, Melbourne, (1999) (5) Wang, B; Xu, D.L.;Xiao, G.X. and Yu, A.B. – Numerical study of gas-solid flow in a cyclone separator Proc. 3rd Intl. Conf. on CFD in Minerals and Process Industries, Melbourne (2003). (6) Gong, A. L. and Wang, Lian-Ze , – Numerical Study of Gas Phase Flow in Cyclones with the Repds, Aerosol Sci. and Technology, 38:5, 506 – 512 (2004) (7) Kegge, S.J – Numerical Simulation of an Axial Hydrocyclone 12 Dec 2000 http://www.ahd.tudelft.nl/downloads/Sander_ Kegge.pdf (8) Ko, J. – Numerical modelling of highly swirling flows in a cylindrical through-flow hydrocyclone Licentiate Thesis, Royal Institute of Technology, Department of Mechanics Stockholm, March (2005) (9) Tomlinson, G.H. II US Patent 3096275 (1961 – continue in part of 1951 filing) (10) Hill, J.; Hoglund, H. and Johnsson, E. – Evaluation of screens by optical measurements Tappi J. 58(10) 120 (1975) (11) Corson, S.R. and Tait, J.D. – Prediction of the performance characteristics of a centrifugal cleaner Tappi J. 60(8) 126 (1977) (12) Pesch, A. W. "Process for separation of fibers having different characteristics", US Patent 3085927 (1963). (13) Paavilainen, L. – The possibility of fractionating softwood sulfate pulp according to cell wall thickness Appita J. 45(5) 319 (1992) (14) Wood, J.R, Grondin, M and Karnis, A. Characterisation of mechanical pulp fines with a small hydrocyclone. Part 1 The principle and nature of the separation J. Pulp Paper Sci 17(1) J1-J5 (1991) (15) Li, M.; Johnston, R.; Xu, L.; Filonenko, Y. and Parker, I. – Characterisation of hydrocyclone (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) separated eucalypt fibre fractions, J. Pulp Paper Sci. 25(8) , 299 (1999) Karnis, A. – Pulp fractionation by fibre characteristics Paperi ja Puu 79(7) 480 (1997) Wang, Z.; Dong, S.; Gartshore, I and Salcudean, M - Fiber separation in hydrocyclones I . Proc. 2002 Tappi Fall Conf. 1 (2002) Park, S.; Venditti, R.A.; Jameel, H. and Pawlak, J.J. – The effect of fibre properties on fibre fractionation using a hydrocyclone J. Pulp Paper Sci. 31(3):1 (2005). Jonas Bergström – Flow Field and Fibre Fractionation Studies in Hydrocyclones Doctoral Thesis , Royal Institute of Technology Stockholm, Sweden (2006) Lepola, I.; Kaila, E. and Ryti, N. – Removal efficiency of a centricleaner, studied with a tracer and activation analysis Papper och Tra Specialnummer 4a 315 (1969) Ferguson, J.W.J. – Theoretical aspects of a pulp suspension flowing in a conventional hydrocyclone Tappi J. 71:125 (1988) Neese, Th.; Duceck, J. and Minkov, L. – Separation of finest particles in hydrocyclones Minerals Engineering 17:689 (2004) Lynch, J and Rao, T.C. – Modelling and scaleup of hydrocyclone classifiers Proc. 11th Intl. Mineral Processing Congress, Cagliari (1989) Vallebuona, G.; Casali, A. and Bevilacqua, P. – Small-diameter hydrocyclones: Performance prediction by empirical models Proc. 7th Intl. Conf. Minerals and Metallurgical Processing. Vallebuona, G.; Casali, A.; Ferrara, G.; Leal, O. and Bevilacqua, P. – Modelling for small diametre hydrocyclones. Minerals Engineering, 8: 3 321 (1995). Arterburn, R.A. Dianne Nielsen [diales@internode.on.net]. The sizing and selection of hydrocyclones Krebbs Engineers http://krebs.com/documents/83_sizing_select_ cyclones.pdf Kawatra, S.K.; Bakshi, A.H. and Rusesky, M.T. – Effect of viscosity on the cut (d50) size of hydrocyclone classifiers Minerals Engineering 9(8) 881 (1996). Rawlins, C.H. – Application of Multiphase Desander Technology to Oil and Gas Production Proc. 3rd Intl. Conf. on Multiphase Technology, Banff, (2002). Gulichsen, J. and Fogelbohm, C-J Papermaking sci. and technology Helsinki (2000). Coulson, J.M. , Richardson, J.F., Blackhurst, J.R. and Harker, J.H. Chemical Engineering Vol 2, 3rd Edn , Pergamon. (1985) Haider, A and Levenspiel, O. – Drag coefficient and Terminal Velocity of Spherical and Nonspherical Particles. Powder Tech. 58 6370 (1989). Heywood, H – The calculation of particle terminal velocities J Imperial Coll. Chem Eng Soc. 4:17-29 (1948). Heiss, J.F. and Coull J. – The effect of orientation and shape on the settling velocity of nonisometric particles in a viscous medium Chem Eng Prog 48(3) 133-140 (1952). Weiss, N.L. (Ed) SME Minerals Processing Handbook Vol 1, Society of Mining Engineers, New York (1985) Kadant Lamort: Kadant Cyclotech General Trade literature on centrifugal cleaners. Vol 62 No 1