Roll compaction of maize powder

Chemical Engineering Science 60 (2005) 3919 – 3931 www.elsevier.com/locate/ces Roll compaction of maize powder Yus Aniza Yusofa,1 , Andrew C. Smithb , Brian J. Briscoea,∗ a Department of Chemical Engineering & Chemical Technology, Imperial College London, London SW7 2AZ, UK b Institute of Food Research, Norwich Research Park, Colney, Norwich NR4 7UA, UK Available online 2 April 2005 Abstract This paper presents a study of a roll compaction process as a dry granulation method for typical food materials such as maize powder. This process is widely applied in industry as it can continuously produce large quantities of granular product at comparatively low cost. The objectives of this work were to predict the roll compaction performance from a simple measurement involving uniaxial die compaction using the classical Johanson model. This involved determination of the optimum operating conditions for the production of granules as evaluated by apparent density. In the current work, a smooth counter-rotating rolling mill with a roller diameter of 0.08 m and a roller width of 0.20 m was used. The operating conditions for the rolling mill are shown to be influenced by parameters such as the roll gap, the roll speed, the feed powder amount, and the friction ratio. Material properties such as the compressibility factor and the angle of wall friction were investigated using uniaxial die compaction. The angle of wall friction was analysed using both contact mechanical and continuum mechanical approaches. The results indicated that this simplified approach can be used to provide a quantitative prediction of the extent of the roll compaction performance, and can be used to design optimal roller geometries and operating conditions. 䉷 2005 Elsevier Ltd. All rights reserved. Keywords: Modelling; Dry granulation; Rolling mill; Maize flour; Uniaxial die compaction; Roll compaction; Friction; Slip-stick behaviour 1. Introduction Granulation is a process whereby small particles are gathered together to form larger cohesive masses in which the original particles can still be identified (Ennis and Lister, 1997). Generally, the subject of granulation has been divided into ‘dry’ and ‘wet’ granulation processes. In dry granulation, dry powders are brought into contact by compression, and examples are roll compaction and uniaxial die compaction. Dry granulation is of particular interest in industry as the final product requires no liquid binder and drying process, and therefore costs less to operate as it requires simpler equipment (Augsburger and Vuppala, 1997; Miller, 1997). It has special advantage for handling of moisture sensitive material. ∗ Corresponding author. Tel.: +44 20 7594 5561. E-mail address: [email protected] (B.J. Briscoe). 1 Current address: Department of Process and Food Engineering, Fac- ulty of Engineering, Universiti Putra Malaysia, 43400 UPM Serdang, selangor, Malaysia. 0009-2509/$ - see front matter 䉷 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2005.02.025 This study considers the roll compaction of maize flour to determine the optimum operating conditions and to predict the roll compaction performance from a simple measurement of uniaxial die compaction using a classical but not fully tested model of a rolling theory for granular solids developed by Johanson (1965a). There has been no reported work on the experimental validation of Johanson’s model for the prediction of the roll force and the roll torque, except work reported by Bindhumadhavan et al. (2004) for the prediction of the pressure profile in the nip region during roll compaction. The influence of parameters such as the roll gap, the roll speed, the feed powder amount, and the friction ratio were investigated. The friction ratio, defined as the ratio of the front roller speed to the back roller speed, influenced the slip-stick behaviour between the powder and the roller surface, and it may also indirectly influence the adhesional and frictional forces which are crucial for the granule formation. Material properties such as the compressibility factor, the effective angle of internal friction and the angle of wall riction were 3920 Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 included in order to test the viability of Johanson’s model. The comparison with the model indicated that this simplified approach may be used to provide a quantitative prediction of the extent of the roller compaction performance. Feed material roll force   Released region 2. Theory 2.1. Uniaxial die compaction and compressibility The material properties of the feed powder are investigated by using uniaxial die compaction which is widely used to determine the mechanical properties of various materials such as metals, ceramics, pharmaceutical materials, and in the present case, food powders (Peleg et al., 1982; Fernando, 1986; Charalambides et al., 2001). The powder is confined between two punches and a cylindrical stainless steel die and is placed on a fixed table. A force is applied at a constant velocity onto the upper punch, while the lower punch does not move within the mechanical assembly. The results from the uniaxial die compaction are presented in the form of the applied pressure–volume relationship, compressibility factor and frictional properties. Compressibility is defined as the ability of the powder to reduce in volume caused by the application of the applied pressure. It is widely presented in the form of a compressibility factor K, obtained from an empirical relationship between the bulk density and applied pressure in the high pressure region and is referred to as a tablet material law (Johanson, 1965a)
K
1 1 , (1) =
2 2 where
1 and
2 are the maximum and minimum pressures, respectively, 1 and 2 are the corresponding bulk density of the material from the maximum and minimum pressures, respectively. 2.2. Effective angle of internal friction Internal or interparticle friction exists whenever particles touch one another. By using a shear cell test, the interparticle friction is characterised by a yield locus at maximum shear stress under any normal stress and the angle below the yield locus is the angle of internal friction (Jenike and Shield, 1959). This concept was extended from the Mohr–Coulomb criterion (Jenike and Shield, 1959) and has been simplified by Johanson (1965a) as is shown in Fig. 1b. The combination of the yield loci defines an effective yield locus, and an effective angle of internal friction, , which indicates the resistance of the powder during flow (Jenike and Shield, 1959). The flowability of the powder decreases as the angle  increases. A typical value of  for most powders is between 25◦ and 70◦ (Thomson, 1984). However, there is no direct method of estimating the effective angle of internal friction  for a cohesive powder except by using a shear cell test (Tüzün, 1987). Roll torque Slip region Nip region Roll diameter roller roll gap nip angle (a) granules feed angle Shear Effective yield locus Wall yield locus stress  A N  2 w O
2 (b) M

1 Normal stress
n Fig. 1. (a) Schematic presentation of a roll compaction process. (b) Effective angle of internal friction and angle of wall friction (Johanson, 1965a). 2.3. Angle of wall friction Wall friction exists whenever the particles are in contact with wall surfaces. It is a measure of resistance of powder to flow along the wall surface (Thomson, 1984). In a shear cell test, when shear is applied to the powder at a constant shear rate, friction will develop between the powder and the wall surface, and is presented as a wall yield locus and the angle below is called an angle of wall friction w as shown in Fig. 1b (Johanson, 1965a), related to the coefficient of wall friction, , by: = tan w, (2) There are three main approaches to interpret the angle of wall friction; continuum mechanics, kinematics, and contact mechanics approaches (Tüzün, 1987). In this paper, both continuum and contact mechanical approaches are used, which are based upon the transmission of stresses in the die and are described in the equation (Macleod and Marshall, 1977):
=
yy + w , (3) where
is the applied pressure,
yy is the transmitted pressure, and w is the interfacial shear stress. The interfacial shear stress will largely determine the magnitude of the wall friction. Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 2.3.1. Continuum mechanics approach: a differential slice technique Walker usefully modified Janssen’s approach to write the transmission of stresses as (Janssen, 1895; Walker, 1966):

yy H ln = −4 Kw FD , (4)
D where
is the applied pressure,
yy is the transmitted pressure, Kw is a factor where the magnitude depends upon the equilibrium state of material, FD is the distribution function, H is the current height of the cylindrical compact, and D is the diameter of the cylindrical die. In order to obtain the estimate of the coefficient of the wall friction, , ln (
yy /
) is plotted against aspect ratio (H /D) with the constant Kw = 0.4 (Briscoe and Evans, 1991) and FD is unity (Briscoe and Evans, 1991). nip angle, which depends upon the physical properties of the feed powder, the nature of the feeding mechanism and the physical characteristics of the rolls (Funakoshi et al., 1977). Problems such as aeration and a rapid increase in densification usually occur in the nip region (Pietsch, 1976; Funakoshi et al., 1977). Below the slip and the nip regions is the release region, where the influence of the elasticity of the powder is apparent (Johanson, 1973). (5) 2.4.2. Slip region Johanson (1965a) used the Jenike and Shield (1959) yield criterion to describe the powder behaviour under continuous shear deformation by assuming the powder to be isotropic, cohesive, and compressible. The powder characteristic in the slip region is determined by the effective angle of internal friction and angle of wall friction. The effective angle of internal friction  represents the plane-strain condition of the powder between the rolls and the angle of wall friction w represents the relationship of the shear (tangential) and normal forces at the roll surface (Johanson, 1965a). This concept has been developed by Jenike and Shield (1959) based on the Mohr–Coulomb criterion and has been simplified by Johanson (1965a) as is shown in Fig. 1 where  is the angular position of the roll “bite”; essentially the roller traction angle. The shear stress and the normal stress at the roll wall surface are then represented by a feed angle  (Johanson, 1965a). (6) 2 = 2.3.2. Contact mechanics approach: an adhesion theory of friction According to the adhesion theory of friction, friction arises from two distinct sources, namely adhesion and deformation (Bowden and Tabor, 1954). The friction force F, at the die walls is mainly caused by the adhesive friction from the solids that are dissipated in an interfacial shear zone by the interfacial shear stress w , and at a specific real area of contact A, and is written as (Bowden and Tabor, 1954) F = w A. 3921 The friction force (Briscoe et al., 1985) is: ′ F = kW m . The coefficient of wall friction 1991): F ′ = = kW m −1 , W − arc sin
sin w sin   − w . (8) is (Briscoe and Evans, (7) where k is the friction factor, W is the applied force, and m′ is the applied force index. 2.4. Roll compaction The successful development of a dry granulation process using the roll compaction greatly depends upon the roll dimensions, the powder behaviour, and the roll operating parameters (Johanson, 1965a). 2.4.1. Powder behaviour The transport behaviour of the powder during the roll compaction process is described based on the slip and the nip regions (see Fig. 1) (Johanson, 1965a). In the slip region, sliding occurs between the powder and the roll surface as the powder is predensified and forced in between the rolls. In the nip region, the powder is compressed and rearranged, and densification occurs as well as plastic deformation (Pietsch, 1976). The extent of the nip region is characterised by a The feed angle is the angle of delivery that depends upon the width of the feed opening above the rolls (Pietsch, 1976). 2.4.3. Nip region In the nip region, the powder behaviour has been postulated to obey the tablet material law from a simple uniaxial die compaction test (Johanson, 1965a). From the applied pressure–volume relationship, the compressibility factor K of the powder may be obtained at high applied pressure, and this value is very important for design purposes. If the ratio of the roll gap to the roll diameter Rg /Dr is much less than 1 (in this paper, Rg /Dr is between 0.0013 and 0.0044), and the roll surface used is smooth, the nip angle depends greatly upon the powder properties K,  and  (Johanson, 1965a). 2.5. Roll operating parameters and roll design Roll operating parameters such as roll gap, roll speed, feed powder amount, and friction ratio, as well as roll design such as roll diameter, roll force, and roll torque influence the performance of a rolling mill or a roll compaction process. 3922 Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 2.5.1. Roll diameter The roll diameter Dr is a fixed parameter that cannot be readily adjusted to provide different rolling conditions. The maximum pressure exerted by the rollers is greatly influenced by the roll diameter and is given as (Johanson, 1965b):
1 − sin  1 − f Pmax = R1 P0 1 + sin  d + Rg K (9) ×[Dr (1 − cos ) cos  + d + Rg cos ] , where R1 is the log pressure ratio, P0 is the estimated feed pressure applied at the roll bite, d is the pocket depth for briquette types of rolls and may be assumed as zero for smooth rollers, f is the recycle fraction (ratio of the powder lost from the roll bite to the feed rate), Rg is the roll gap,  is the nip angle,  is the effective angle of internal friction and K is the compressibility factor. The feed pressure at the roll bite P0 (psi) for a rolling mill with a gravity feeding system is given by Johanson (1965b) as: P0 = 0.40 l, 2.5.2. Roll force In practice, the roll separating force Rsf is used to measure the performance of a roll compaction process and is written as: Rf , l (11) where Rf is the average of the forces generated from the front and the back roll, and l is the roll width. Johanson (1965a) has written the roll force as a first-order model and is given as: Rf =
lD r Fr , 2 (12) where
is the applied pressure that may produce good quality of granules, l is the roll width, Fr is the force factor that may be obtained from Johanson (1965a) as a function of (d + Rg )/Dr at different compressibility factors, K. 2.5.3. Roll torque The roll torque Tq generated from the roll compaction is used to calculate the coefficient of sliding wall friction r , which later will determine whether slip-stick occurs (see Section 2.6). The coefficient of sliding wall friction r may be calculated by: r =  Rf = Tq , R (14) where R is the roll radius. Roll torque Tq may also be estimated by the integration of the moments about the roll axes of the constant horizontal compaction forces (Pietsch, 1976) and is given by Johanson (1965a) as: Tq =
lD 2r T , 8 (15) where T is the torque factor that may be obtained from Johanson (1965a). The roll power Rp (W ) is then written as (Johanson, 1965b): Rp = 2.796 × 10−3 Pmax lD 2r T S r , (16) where Sr is the roll speed (rpm) and Pmax is the maximum exerted pressure (psi). (10) where 0 is the initial bulk density of powder (lb in−3 ) before the roll compaction process and l is the roll width (in). Roll separating force Rsf = for shear force 
r Rf , where Rf is the roll force, and  is the shear force which may be obtained from the equation given below as: (13) 2.5.4. Roll speed The roll speed Sr mainly determines the pressing time (retention time) of the powder that passes through the nip region, which is related to the amount of material passing through the nip or the throughput Cc (kg h−1 ) and is written as (Pietsch, 1991): Cc = 60g Dr lR g Sr , (17) where Dr is the roll diameter (cm), l is the roll width (cm), Rg is the roll gap (cm), Sr is the roll speed (rpm) and g is the apparent density (kg cm−3 ). The pressing time in the roll mill tp was described as (Pietsch, 1991): tp = Fr , 2 Sr (18) where Fr is the force factor. 2.6. Slip-stick behaviour of powder In a roll compaction process, the speed of the front roll and the back roll are usually similar. A slight variation of the roll speed may cause slip to occur which depends upon the material properties of the powder (Pietsch, 1976). If the coefficient of static friction exceeds the coefficient of sliding friction of the powder, the powder may exhibit slip-stick behaviour. However, if the coefficient of the sliding friction exceeds the coefficient of static friction, continuous slip motion will occur. When slip occurs, the roll will move faster than the powder (Schönert and Sander, 2002), which induces extrusion into the process and wear on the roll surfaces (Lim et al., 1997). Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 3. Experiment 3.1. Feed powder Maize flour (Fine Polenta F) was obtained from Smiths Mills, Worksop, UK. The dry matter composition of the maize was 80.5% starch, 13% protein, 1.7% dietary fibre, 1.4% fat, 0.7% free sugar and 0.2% ash. The material properties before the granulation process are shown in Table 1. The mean particle size was measured by using a laser particle size analyser, Mastersizer 2000 (Malvern Instruments Ltd., UK). The bulk density, b , was measured by filling a tare of 100 ml graduated cylinder (diameter of the cylinder is 30 mm) to approximately 70 ml mark and the weight recorded. Then, the tapped density, t , was measured after 500 taps (until the powder height in the container was constant). A pycnometer, AccuPyc 1330 device (Micromeritics Instrument Corp., Norcross, GA) was used to determine the true density, s , by utilising helium gas as the displacement fluid (Pietsch, 2002). All of the measurements were carried out three times. The inter-relationships between the porosities and densities are given by Ennis and Lister (1997). Large values of the Carr Indices and Hausner Ratios indicate poor flowability and they are generally correlated with high values of the interparticle friction (Carr, 1965; Hausner, 1967), which usually correspond to a large effective angle of internal friction. 3.2. Uniaxial die compaction A commercial die compaction machine (Lloyds, UK) with a cylindrical stainless steel die (Specac, UK) of a diameter 20±0.5 mm was used, the roughness of the components was comparable to that of the surface roughness of the rolling mill. The uniaxial die compaction machine comprised a fixed bottom punch and a moving top punch. At the upper punch, an applied force transducer (Sensotec Inc., USA) with a maximum allowable force of 50 ± 1 kN and a displacement transducer were installed. The displacement and the applied force exerted during the compaction process were recorded on a computer. Another force transducer (RDP Electronics, UK) was connected with an additional computer to measure the transmitted force (see Eq. (3)). The maximum allowable force for the transmitted transducer was 90 ± 2 kN. Firstly, 3.00±0.01 g of the maize flour was poured gradually using a plastic funnel into the cylindrical stainless steel die. The die was then tapped 20 times on a surface of a bench for data consistency and in an attempt to form a homogeneous density distribution. Next, the transmitted force transducer was placed at the bottom of the die and 10 ±1 kN force was applied from the upper punch with a crosshead speed of 0.5 mm s−1 . This process was called the ‘loading’ process. After the desired force was reached the ‘unloading’ process occurred where the applied force was removed from the upper punch. The data of the applied force corre- 3923 sponding to the displacement of the compacted maize were logged by a computer. In order to release the compacted powder from the die, the bottom punch was removed and the upper punch was consolidated. The thickness of the compacts was measured by using a 25 mm External Micrometer (Vernier, England). Compression tests were repeated for the applied forces of 20, 30, 40, 45 and 49 kN. The 49 kN applied data were used as input for testing the Johanson (1965a) model. 3.3. Roll compaction Experiments for roll compaction were carried out on a laboratory rolling mill (Francis Shaw, Manchester, UK) which was equipped with smooth rollers of 0.08 m (80 mm) diameter and 0.20 m (200 mm) width (the rolls’ surface finish was less than 0.1
m centre line average) that were horizontally arranged. In order to measure the roll compaction conditions the rolling mill was instrumented. Two compression type load cells or also known as force transducers with a Direct Current (DC) Bridge, with a maximum allowable force of 9071.85 N (2000 lb) were connected to the left and right sides of the moving front rolls in order to measure the magnitude of the roll separating forces. The signals of each of the transducers were transmitted by means of a high precision slip ring to transducer amplifiers, model S7M-Z (R.D.P. Electronics Ltd, UK) to meet the conditioning needs of the type of the DC Bridge transducers. The signal of the transducers was connected to a computer via a direct memory access A/D converter card. A torque transducer was used to measure the magnitude of the torque transmitted to the rolls. The drive shaft at the back roll, was fixed to transmit signal of the torque generated during rolling with the maximum allowable torque of 50 Nm (433 lb in). The torque signal was transmitted by a slip ring transducer amplifier, PR 9307 Carrier Frequency Bridge (Phillips, Germany) via a direct memory access A/D converter card to the same computer connected to the force transducers amplifier. The rolling mill was equipped with a dual drive system, with a maximum power of 1.1 kW using an AC Vector Controlled motor. Gear boxes transmitted the drive via a reduction gearbox and a belt drive system. A control system consisted of the drive shaft, the emergency and reset push buttons, and roll speed and friction ratio thumbwheel controls. The feed powder was fed between the rollers by a gravity feeding chute with a constant feeder height of 0.2 m. An amount of feed powder ranging from 0.03 to 0.16 kg was used. Variations were made in the roll gap, the roll speed, the feed powder amount and the friction ratio (ratio of the front roll speed to the back roll speed). The compacted maize powder was stored in an airtight plastic bag for later characterisation. In this study, the maximum allowable roll separating force of the force transducer was 0.45 kN cm−1 , using the standard roll mill classification. 3924 Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 Table 1 Material properties of maize flour Material properties (×10−6 /m) Mean particle size dp Bulk density b (kg m−3 ) Tapped density t (kg m−3 ) True density s (kg m−3 ) Bulk porosity b Apparent density g (kg m−3 ) Moisture content Mc (%) Carr Index CI (%) Hausner Ratio HR Values References 135 ± 15 446.25 ± 23.50 697.60 ± 18.25 1504.7 ± 16.0 0.475 ± 0.003 850.0 ± 5.5 11.1 ± 1.0 39 (poor flowability) 1.6 (poor flowability) Carr (1965) Hausner (1967); Hayes (1987) 3.4. Granule characterisation The efficiency of the granule formation during roll compaction was characterised by the apparent granule density g , which includes closed pores and open pores (Hancock et al., 2003), and is defined as the ratio of powder mass to the envelope volume. The apparent density was measured by an envelope density analyser, GeoPyc 1360 (Micromeritics Instrument Corp., Norcross, GA), with an accuracy of ±1.1%. 4. Results 4.1. Compressibility factor From the measured inter-relationship between the bulk density and the applied pressure, the compressibility factor K was calculated (Johanson, 1965a), (Eq. (1)). The value of the compressibility factor K for maize powder was found to be 9.03 ± 1.81 with the regression value R 2 of the relationship having a value of 0.9880. The value of K greatly depends upon the material properties. For instance, K for coffee creamer is 12.50, powdered sucrose is 33.3, and powdered salt is 50.0 (Peleg et al., 1982). 4.2. Effective angle of internal friction A value of 52◦ for the effective angle of internal friction  of maize powder was adopted, taken from Fitzpatrick et al. (2004), which was measured with a Jenike shear cell with a true density s of 1490 kg m−3 and a mean particle size of 49
m. Although the mean particle size of the maize flour used in this study was 135
m and is relatively large compared to that of Fitzpatrick et al. (2004), both feeds can be considered as fine powders having particle size less than 170
m under the Geldard Group C particles class, which are difficult to flow (Geldard, 1973). This was supported by the calculation of Hausner Ratio and Carr Index, shown in Table 1, which indicated that the powder was very cohesive and difficult to flow. The characteristic of the maize flour used by Fitzpatrick et al. (2004) was also categorised as difficult to flow based on the flow index of 1.5 and was classi- fied as very cohesive (Tomas and Schubert, 1979). The flow index was defined as the inverse slope of the flow function, from the relationship between the unconfined yield strength of the powder and the major consolidating stress in a Jenike shear cell. In order to use these data in the Johanson (1965b) analysis, the  value should be either 30◦ , 50◦ or 70◦ ; therefore the value of  was taken as 50◦ . 4.3. Angle of wall friction Both continuum mechanics and contact mechanics approaches were used in this paper to obtain the value of angle of wall friction w . The value of w obtained was incorporated later in the established models to predict the roll design parameters. 4.3.1. Continuum mechanics approach: a differential slice technique In order to calculate the coefficient of wall friction , the value of the slope from a plot of the stress transmission ratio as a function of compact aspect ratio was substituted into Eq. (4) (see Section 2.3.1). The calculated value was 0.25 which was based upon the friction between the maize powder and the stainless steel die wall. The angle of wall friction w was thus calculated by using Eq. (2) and was given as 14◦ . The value of is in good agreement with values published for polymers ranging from 0.15 to 0.25 (Briscoe and Evans, 1991). Fitzpatrick et al. (2004) have given the angle of wall friction: for maize powder: 13◦ ( = 0.23), salt 27.3◦ ( = 0.52) and sugar 19.1◦ ( = 0.35). 4.3.2. Contact mechanics approach: an adhesion theory of friction The friction force coefficient decreased as the applied force increased, similar to the trend reported by Briscoe et al. (1985). By fitting Eq. (6), values of k=0.60 and m′ =0.88 were obtained. The value of m′ was between 2/3 (0.67) and unity which indicates that adhesive friction arises from the real area of contact at the interface. The value of m′ obtained is in good agreement with Briscoe et al. (1985) with the value of 0.87 for maize powder. The value of calculated using Eq. (7) was 0.17, and by using Eq. (2), the value of w was equal to 10◦ . Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 4.4. Roll compaction The effects of the roll compaction process as a dry granulation method on the roll operating conditions such as the roll gap, roll speed, feed powder, and friction ratio were investigated. By using a trial-and-error procedure, the optimum operating conditions were also determined. The optimum operating conditions were: roll gap 0.15 mm, roll speed 40 rpm, feed material 0.07 kg, and friction ratio 1.0, as described below. 4.4.1. The effect of the quantity of feed powder The quantity of the feed powder was defined as an amount of material that was poured into the feeder during the start of each run of the rolling mill with the other operating conditions such as the roll gap, roll speed, and friction ratio were kept constant and at their optimum operating conditions. The high values of the roll separating forces indicate that the amounts of the powder used were sufficient to induce extensive powder-wall surface contact. The lowest feed powder amount of 0.03 kg showed rather low values of roll separating forces of 700 N m−1 . However, for amounts of feed powder of 0.14 and 0.16 kg it was found that a large amount of maize powder was accumulated at the front and back of the rollers and collapsed without being compacted, which indicated that, an excessive amount of the feed powder may cause an inefficient feeding of rolling process. During the experiments, it was also observed that the granules produced were in the form of weak thin flakes, of typical mean diameter 10 mm, for the entire range of the feed powder used. However, no apparent slip occurred in any of the feed powder. The highest apparent density occurred at the feed powder amount of 0.07 kg. This particular amount of feed powder may be considered as a sufficient amount of powder to produce high quality relatively tough granules. Therefore, this amount of feed powder was used in the subsequent experimental work to investigate the effect of the roll gap, the roll speed and the friction ratio. 4.4.2. The effect of the friction ratio The friction ratio fr is defined as the ratio of the speed of the “front” roll to the speed of the “back” roll, the identification is arbitrary as there were not uniquely identifiable differences between the rollers. In this study, the effect of the friction ratio was investigated to relate the granule apparent density with the ratio of the roller speeds. The friction ratio influences the slip-stick behaviour of the powder and the roller surface and it may indirectly influence the adhesion and frictional forces, which are crucial for granules formation. A friction ratio of 0.2–2.0 was used with the roll gap, roll speed, and feed powder kept constant at their optimum operating conditions. The highest roll separating force occurred at a friction ratio fr equal to 1. However, at friction ratios less than or more than 1, the performance of the 3925 rolling mill was significantly reduced. This was due to the fact that, as rollers moved at speeds relatively different from each other, this reduced the contact between the powder and the roller surfaces, and hence allowed more slip to occur. However, at a friction ratio of 0.7, the roll separating force was only slightly less than at a friction ratio of 1.0, and the coefficient of sliding wall friction r values sometimes became less than the static coefficient of wall friction = 0.25 of the powder. Slip-stick behaviour then occurred. During the experiment under discontinuous sliding conditions, it was observed that a ‘zebra’ stripe-like surface morphology or a ‘strip’ type appearance formed on the front roller surface at friction ratio of 0.7. The ‘zebra’ morphology surface here means that the powder formed adhered strips of transferred powder on the roller surface with no such deposited powder between the strips. This phenomenon can be tentatively attributed to the presence of air trapped in the nip region due to the higher speed of the back roll causing aeration to the maize powder, which may result in a poor product quality and induce wear on the roll surface. This result is consistent with that reported by Pietsch (1991) where entrapped air was shown to exist within the roll gap, and may effect the feeding of the powder and the product quality. As expected, the apparent density of the maize powder was highest when the friction ratio was 1. 4.4.3. The effect of the roll gap The variations of roll gap between 0.10 to 0.35±0.05 mm were used so that it was within the range of the mean particle size of the maize powder of 135
m. The roll speed, feed powder, and friction ratio were kept constant at their optimum operating conditions. In general, the experimentally measured roll separating force increased as the roll gap was decreased. The smallest roll gap of 0.10 mm (100
m) provided the highest separating force of 1000 N m−1 between the rollers. As the processing time increased, the roll separating force gradually increased until the rolling time t was approximately 45 s and remained constant for approximately 10 s before it decreased very rapidly. The increasing separating force was due to the increasing amount of maize powder sticking to the roll surface as the time increased, which then allowed more contact between the powder and the roll-wall surface, causing more powder cohesion, and hence an enhanced initial formation of granules. Direct evidence of the powder slip-stick behaviour was observed from the relationship between the coefficient of sliding wall friction r and the rolling time. The values of were calculated using two approaches to be 0.17 and 0.25, respectively. Slip occurs when r exceeds of the powder. For roll gaps of 0.10–0.20 mm, slip only occurred for the first few seconds of the rolling time, indicated by r values larger than = 0.25. However, this condition of slip may be ignored due to the initial unsteady state of the process. For roll gaps of 0.25–0.35 mm, on the other hand, slip occurred Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 during the whole compaction process, which may be caused by the large roll gap where the low separating force did not facilitate effective powder contact with the roll surface. From the observation made during the experiment, there were some granules produced in the form of thin flakes with a mean size of approximately 5 mm (0.005 m). However, these flakes broke when in contact with subsequently compacted maize powder. The weak inter-particle bonding caused these granules flakes to break very easily. Therefore, the mean granule size may not be used to characterise the process. Weak granules are still acceptable when an instantaneous characteristic, such as dissolving easily in water, is required and this concept also may be applied in the pharmaceutical industry as an intermediate process before tabletting. The selection of the granulation process depends on the product application, as extensively explained by Capes (1980). The reason for the weak granule formation was because the forces applied to the rolls were too small to produce a coherent compact. The maximum separating force that was generated during the experiment was 45 kN m−1 (0.45 kN cm−1 ), which may be considered as extremely low, compared to the roll separating force used by Bultman (2002) to compact microcrystalline cellulose of 7 kN cm−1 ; about 15 times higher than the roll separating force in this study. Moreover, due to the viscoelastic characteristic of the maize powder, the low separating force applied may have allowed significant elastic deformation compared to the plastic deformation to occur. Therefore, large stored elastic strains within the powder prevented the permanent formation of cohesive junctions (plastic junctions) between the particles. The most robust parameter used to characterise the effectiveness of the roll compaction process was the apparent density of the compacted maize powder, the accuracy of the measurements was approximately 2%. The highest apparent density measured was 980 kg m−3 at a roll gap of 0.10 mm (100
m), and decreased gradually to 900 kg m−3 as the roll gap increased to 0.35 mm (350
m). These results clearly indicate that the granulation process occurred during roll compaction as there were increases of the apparent density from the uncompacted apparent density of 850 kg m−3 . For roll gaps of 0.10–0.20 mm, the increase was about 15%. For the subsequent investigations, the roll gap of 0.15 mm was chosen because smaller gaps often caused a small fraction of size reduction to occur, which is beyond the scope of this work. Therefore, the optimum roll gap was taken as approximately equal to the mean particle size of the maize powder (135
m). 4.4.4. The effect of the roll speed The roll speed ranging from 10 to 60 rpm was used (see Fig. 2) with the other operating conditions of roll gap, feed powder, and friction ratio were kept constant at their optimum operating conditions. The roll separating force in- Roll separating force Rsf / Nm-1 3926 10000 8000 Roll speed Sr / rpm 10 20 30 40 50 60 6000 4000 2000 0 0 20 40 Rolling time t/s 60 Fig. 2. Roll separating force, Rsf as a function of rolling time, t for different roll speeds, Sr at ambient conditions with the feed powder amount Wt of 0.07 kg, the roll gap Rg of 0.15 mm and the friction ratio fr of 1.0. creased as the roll speed increased from 10 to 60 rpm. Less time was required to achieve the maximum roll separating force as the roll speed increased, ranging from 35 s at 60 rpm to 60 s at 30 rpm, and the longest periods being at 10 and 20 rpm. In Fig. 2, the maximum roll separating force can be seen clearly for the roll speeds of 30–60 rpm. Higher rolling times are not shown in this figure as similar magnitudes of the roll forces were achieved after a rolling time of 55 s except for 30 rpm. However, after 60 s for 30 rpm, the roll separating force decreased gradually. The roll separating forces for 10 and 20 rpm, did not show any significant maximum value even though the experiment was carried out for longer than 60 s. Therefore, the maximum separating force was considered to be relatively constant at 1700 and 2000 N m−1 , respectively. Note also that the amount of the material that passed through the roll gap depended upon the roll speed. A higher roll speed meant that more powder passed through the roll gap, therefore leaving very little material on the roller surface. During the experiment, it was also observed that powder leakage occurred at both ends of the rollers and increased as the roll speed increased. As the roll was moving faster, more powder was pushed to both ends of the roller as it passed through the nip region. It was also found that the granules produced were in the form of weak thin flakes as was reported in the previous section. The number of the roll separating force fluctuations depended upon the roll speed. For instance, if 10 rpm was used, 10 fluctuations of the roll separating force may be seen during 10 rotations. The fluctuations were most certainly caused by imperfect alignment of the rollers. The shapes of the transducer outputs during one rotation, may be used to describe the pressure distribution in the nip region. This shape was found to be similar to that given previously by Johanson (1973), Schönert and Sander (2002), and Dec et al. (2003). Extensive roller/powder slip only occurred at the lowest roll speed of 10 rpm. The slow rolling process allowed sufficient time for the powder to slip through the roll gap. At Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 3927 3000 8000 Calculated bulk porosity 1 Calculated bulk porosity 2 6000 4000 2000 0 0.05 0.15 0.25 0.35 Measured Calculated Roll force Rf /N Roll force Rf / N Measured 2000 1000 0.45 Roll gap RgX10-3 /m speeds higher than 20 rpm, there were shorter retention times and hence no slip occurred. Moreover, the contact between the powder and the roll wall surfaces may be increased as the roll speed increased. However, the high roll speeds (50 and 60 rpm) of the rolling process prevented the powder from interacting with the roller walls effectively. This corresponded to the highest apparent density being obtained at the roll speed of 40 rpm, which was used in the subsequent experiments. 0 20 40 Roll speed Sr / rpm 60 80 Fig. 4. Roll force as a function of roll speed. Error bars are based upon six measurements. 3000 Measured Calculated Roll force Rf / N Fig. 3. Roll force as a function of roll gap for bulk porosity 1 ( b =0.475) and bulk porosity 2 ( b = 0.435). Error bars are based upon six measurements. 0 2000 1000 4.5. Model verification 0 0 50 100 150 200 Feed powderWt X10-3/kg Fig. 5. Roll force as a function of feed powder. Error bars are based upon six measurements. 3000 Roll force Rf /N The results of the roll force and the roll torque obtained from the experimental work were compared with the predictions obtained using Eqs. (12) and (15), by incorporating the compressibility factor, the effective angle of internal friction, the angle of wall friction, and the applied pressure–volume relationship from the uniaxial die compaction. The angle of wall friction value was taken as 14◦ , based on the continuum mechanics approach. Figs. 3–6 show the experimental and computed roll force values as functions of roll gap, roll speed, feed powder and friction ratio, respectively. The error bars of the measured roll force were based upon a series of 6 measurements. Figs. 7–10 show roll torque as a function of roll gap, roll speed, feed powder and friction ratio, respectively. The measured roll force and roll torque are within the range of 200–2500 N and 0.6–1.6 Nm respectively, which are in agreement with finite element simulation results from Dec et al. (2003) for microcrystalline cellulose ranging from 460 to 760 N and 0.96–1.4 Nm, respectively, with the coefficient of wall friction of 0.35. The roll force and the roll torque were calculated based on the applied pressure–volume relationship obtained from the uniaxial die compaction data. The bulk density was calculated for every applied pressure, and the apparent density value calculated with b = 0.475 (bulk porosity 1) from Measured Calculated 2000 1000 0 0.0 0.5 1.0 1.5 Friction ratio fr 2.0 2.5 Fig. 6. Roll force as a function of friction ratio. Error bars are based upon six measurements. Table 1. Therefore, a series of apparent density values was obtained. The roll force and the roll torque were calculated using Eqs. (12) and (15) with Fr = 0.0215 and T = 0.0009 3928 Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 2.0 Measured Calculated Measured Calculated 3.0 Roll torqueTq /Nm Roll torque Tq / Nm 4.0 2.0 1.0 0.0 0.05 1.5 1.0 0.5 0.0 0.15 0.25 0.35 0 0.45 50 Roll gap RgX10-3/m Fig. 7. Roll torque as a function of roll gap. Error bars are based upon six measurements. 100 150 200 Feed powder Wt X10-3 /kg Fig. 9. Roll torque as a function of feed powder. Error bars are based upon six measurements. 2.0 Measured Calculated 2.0 Roll torqueTq /Nm Roll torqueTq /Nm Measured Calculated 1.5 1.0 0.5 1.5 1.0 0.5 0.0 0.0 0.0 0 20 40 Roll speed Sr /rpm 60 80 Fig. 8. Roll torque as a function of roll speed. Error bars are based upon six measurements. from Johanson (1965a) for a roll gap of 0.15 mm and K = 9.03, obtained from the uniaxial die compaction data. For a comparison between the roll force and the roll torque, the calculated apparent density was extracted to match the apparent density from the roll compaction experiment. For instance, at a roll gap of 0.15 mm, and with the other operating conditions kept constant at their optimum operating conditions, g = 971.68 kg m−3 ; the same values of g were extracted from the uniaxial die compaction data. The calculated roll force and roll torque values were 1.9 kN and 1.59 Nm at these particular conditions and can be compared favourably to the measured values of 1.8 kN and 1.60 Nm, respectively. The measured roll force was obtained from the maximum separating force (Eq. (11), also see Fig. 2) and roll torque was fairly constant throughout the rolling process. The magnitude of the roll force and the roll torque depended upon the roll gap, the roll diameter, the roll width and the compressibility factor. In this paper, the roll gaps was 0.5 1.0 1.5 Friction ratio fr 2.0 2.5 Fig. 10. Roll torque as a function of friction ratio. Error bars are based upon six measurements. varied. Other spreadsheets were prepared for the roll gaps of 0.10, 0.20, 0.25, 0.30 and 0.35 mm. Finally, the calculated roll force and roll torque were then compared with the results measured from direct experiment. 5. Discussion 5.1. Comparison of roll force Generally, the value of the calculated roll force agreed quite closely with the measured roll force when investigating the effects of the roll speed (Fig. 4), the feed powder amount (Fig. 5), and the friction ratio (Fig. 6). However, the quality of the agreement reduced at the feed powder amounts of 0.03, 0.14 and 0.16 kg. Too small or too large amounts of feed powder reduced the product quality, and hence the calculation of the roll force was much higher or lower than measured. This observation is in agreement with that of Pietsch (1991) that only a critical amount of feed powder is suitable to produce a good quality of product. For Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 the effect of the roll gap (Fig. 3), the calculated roll force agreed well with the measured roll force at the roll gap of 0.10 and 0.15 mm only. At roll gaps of 0.20–0.35 mm, the calculated roll forces were predicted to be twice the measured roll force. The larger gaps may have allowed the powder to slip, and consequently reduced the contact between the powder and the roll surface. Therefore, the calculated value is expected to deviate since Johanson (1965a) assumed that no slip occurred within the nip region. The calculated and measured roll forces agreed quite well at the optimum operating conditions: the roll gap of 0.15 mm, the roll speed of 40 rpm, the feed powder amount of 0.07 kg, and the friction ratio of 1.0. The Johanson (1965a) model may be used to predict the roll force at different operating conditions, the accuracy of the roll force measurement being approximately 20%. The large range of the roll forces measured may arise as a result of the fine powder which stuck on the roll force transducers, which could have reduced the reading. This effect has also been noted by Schönert and Sander (2002) for the prediction of the shear and pressure as a function of rotational angle in a high-pressure rolling mill (with maximum applied stress of more than 300 MPa). 5.2. Comparison of roll torque The trends observed in the calculated values of the roll torque agreed quite well with the measured values of the roll torque for the effects of the roll speed (Fig. 8), the feed powder amount (Fig. 9) and the friction ratio (Fig. 10). However, the calculated value of the roll torque deviated slightly at the roll speed of 10 rpm, feed powder amounts of 30, 140, and 160 g, and the friction ratios of 0.2, 0.5, and 2.0. For the effect of the roll gap, there were large discrepancies between the measured and the calculated roll torque values, particularly at the roll gap of 0.10 mm (Fig. 7). These discrepancies may be due to aeration in the nip region, which may effect the feeding of the powder, and consequently the measured roll torque was lower than anticipated. Pietsch (1976) noted that aeration may affect the roll compaction process. However, the roll torque agreed quite well for the roll gaps of 0.15–0.30 mm. Among the various important parameters that contributed significantly to the numerical predictions was the compressibility factor obtained from the tablet material law. The application of the tablet material law from the uniaxial die compaction has been questioned in the literature by Sommer and Hauser (2002). There may be large discrepancies between the measured and calculated values used to describe the pressure distribution within the nip region. In general, the torque prediction may be used to provide an estimation of the quantity for a roll compaction process. However, further study is required to improve the quality of such predictions, which has also been noted by Pietsch (1991), and Sommer and Hauser (2002). Sommer and Hauser (2002) have also suggested that a more advanced 3929 Table 2 Data analysis Parameter Values (◦ ) Feed angle  Roll power Rp (W) Throughput Cc (kg s−1 ) Pressing time tp (s) 74–77 0.002–0.044 0.001–0.010 0.003–0.020 modelling method is required for a better prediction of the overall roll compaction performance. 5.3. Bulk porosity effect on the model validation The effect of the initial bulk porosity on the prediction of the model has been investigated and is shown in Fig. 3. At initial an bulk porosity of 0.435 (bulk porosity 2), the roll force curve is three times higher than the roll force at a bulk porosity of 0.475 (bulk porosity 1). Therefore, it is crucial to measure the bulk porosity precisely before the granulation process. The effect of initial bulk porosity is important in powder rearrangement and densification during the entire compaction granulation process, and therefore, a slight change in the initial bulk porosity has a large effect upon the whole compaction and granulation system. The magnitude of the tapped density (Table 1) indicates possible lower values of the bulk porosity. 5.4. Effect of operating conditions Table 2 shows the ranges of feed angle, roll power, throughput, and pressing time calculated from Eqs. (8) and (16)–(18), respectively, for all of the operating conditions. The feed angle, also known as an angle of delivery of the powder, is affected by the roll gap, the roll speed, the feed powder amount, and the friction ratio. It was also influenced by the powder properties, particularly the angle of wall friction w . The roll power decreased as the roll gap and the roll speed decreased. At the roll gap of 0.10 mm, the roll power was relatively high at 0.044 W. Otherwise the roll power remained constant at 0.009 W for changes in feed powder and friction ratio. The througel hput also increased linearly as the roll gap and the roll speed increased but remained constant at approximately 0.005 kg s−1 for changes in feed powder amount and friction ratio. The pressing time increased as the roll gap increased and the roll speed decreased. The pressing time decreased from 0.020 s at 10 rpm to 0.003 s at 60 rpm, and increased from 0.004 s at 0.10 mm to 0.006 s at 0.35 mm, but remained constant at 0.004 s for changes in the feed powder and the friction ratio. In summary, the calculation of roll power, throughput, and pressing time only showed significant changes due to variations in the roll gap and the roll speed. However, the calculated values are relatively low and remain constant for changes in the feed powder amount and the friction ratio. 3930 Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 6. Conclusions The roll compaction parameters such as the roll force and the roll torque have been predicted using Johanson’s model by incorporating simple measurements obtained from uniaxial die compaction. The predictions of the model were in reasonable agreement with the experimental results, and were rather sensitive to the variations in the initial bulk porosity. This model may also be applied where there is a slight deviation when slip occurs in the compaction granulation system. The effects of the roll operating conditions on the roll power, the throughput, and the pressing time were also calculated. The roll gap and roll speed had significant effects on the roll power, throughput, and pressing time, while the feed powder amount and the friction ratio were constant, and in comparison, uninfluential. Dry granulation can be carried out on maize powder using a roll compaction process and the optimum operating conditions, which depend on the material properties and roll design, were determined. This study has demonstrated that the roll compaction performance can be predicted quantitatively, and indicates the necessary protocols for the design, the optimal geometries and the operating conditions for compaction and granulation. Acknowledgements The authors would like to acknowledge Professor Michael J. Adams from Unilever and Dr. Kendall Pitt from Merck Sharp and Dohme, Professor Chris J. Lawrence and Dr. Omar Matar from Imperial College London for the invaluable discussions, and Mr. Dan Parsonage for technical assistance. One of the authors (Y.A.Y.) is also grateful for the scholarship from Universiti Putra Malaysia, Malaysia and A.C.S. was funded from the BBSRC competitive strategic grant. References Augsburger, L.L., Vuppala, M.K., 1997. Theory of granulation. In: Parikh, D.M. (Ed.), Handbook of Pharmaceutical Granulation Technology. Marcel Dekker, New York, pp. 7–9. Bindhumadhavan, G., Seville, J.P.K., Adams, M.J., Greenwood, R.W., Fitzpatrick, S., 2004. Roll compaction of pharmaceutical excipients: experimental validation of rolling theory for granular solids. Private communication. Bowden, F.P., Tabor, D., 1954. Friction and Lubrication of Solids. Oxford University Press, London. Briscoe, B.J., Evans, P.D., 1991. Wall friction in the compaction of agglomerated ceramic powders. Powder Technology 65, 7–20. Briscoe, B.J., Fernando, M.S.D., Smith, A.C., 1985. The interfacial friction of compacted maize powders. Journal of Physics D: Applied Physics 18, 1069–1085. Bultman, J.M., 2002. Multiple compaction of microcrystalline cellulose in a roller compactor. European Journal of Pharmaceutics and Biopharmaceutics 54, 59–64. Capes, C.E., 1980. Particle size enlargement. In: William, J.C., Allen, T. (Eds.), Handbook of Powder Technology. Elsevier Scientific Publishing Company, The Netherlands. Carr, R.L., 1965. Evaluating flow properties of powders. Chemical Engineering 72, 163–167. Charalambides, M.N., Goh, S.M., Lim, L., Williams, J.G., 2001. The analysis of the frictional effect on stress-strain data from uniaxial compression of cheese. Journal of Material Science 36, 2313–2321. Dec, R.T., Zavaliangos, A., Cunningham, J.C., 2003. Comparison of various methods for analysis of powder compaction in roller press. Powder Technology 130, 265–271. Ennis, B.J., Lister, J.D., 1997. Principles of size enlargement. In: Perry, R.H., Green, D.W. (Eds.), Perry’s Chemical Engineers’ Handbook, seventh ed. Mc Graw Hill, New York, pp. 20–56: 20–89. Fernando, M.S.D., 1986. Traction induced compaction of maize powder. Ph.D. Thesis, Imperial College of Science, Technology and Medicine, London. Fitzpatrick, J.J., Barringer, S.A., Iqbal, T., 2004. Flow property measurement of food powders and sensitivity of Jenike’s hopper design methodology to the measured values. Journal of Food Engineering 61, 399–405. Funakoshi, Y., Asogawa, T., Satake, E., 1977. Use of a novel roller compactor with a concavo-convex roller pair to obtain uniform compacting pressure. Drug Development and Industrial Pharmacy 3 (6), 555–573. Geldard, D., 1973. Types of gas fluidization. Powder Technology 7, 285–292. Hancock, B.C., Colvin, J.T., Mullarney, M.P., Zinchuk, A.V., 2003. The relative density of pharmaceutical powders, blends, dry granulations and immediate release tablets. Pharmaceutical Technology April, 64–80. Hausner, H.H., 1967. Friction conditions in a mass of metal powder. International Journal of Powder Metallurgy 3 (4), 7–13. Hayes, G.D., 1987. Food Engineering Data Handbook. Longman Scientific & Technical, New York. Janssen, H.A., 1895. Versuche über Getreidedruck in Silozellen, Zeitschr. d. Vereines deutscher Ingenieure 39, 1045. Jenike, A.W., Shield, R.T., 1959. On the plastic flow of Coulomb Solids beyond original failure. Journal of Applied Mechanics, 26 Transactions of the ASME 81, series E, 599–602. Johanson, J.R., 1965a. A rolling theory for granular solids. Transactions of the ASME Dec., 842–848. Johanson, J.R., 1965b. Factors influencing the design of roll-type briquetting presses. Proceedings of 9th Biennial Conference of International Briquetting Association, pp. 17–31. Johanson, J.R., 1973. Predicting limiting roll speeds for briquetting presses. Proceedings of 13th Biennial Conference of International Briquetting Association, pp. 89–99. Lim, W.I.L., Campbell, J.J., Tondo, L.A., 1997. Chapter 36. Extrusion effects in the high pressure grinding rolls. In: Kawatra, S.K. (Ed.), Comminution Practices. SMME, Littleton, USA, pp. 293–301. Macleod, H.M., Marshall, K., 1977. The determination of density distributions in ceramic compact using autoradiography. Powder Technology 16, 107–122. Miller, R.W., 1997. Roller compaction techonology. In: Parikh, D.M. (Ed.), Handbook of Pharmaceutical Granulation Technology. Marcel Dekker, New York, pp. 99–102. Peleg, M., Moreyra, R., Scoville, E., 1982. Rheological characteristics of food powders. Food Process Engineering. A.I.Ch.E. Symposium Series 78 (218), 138–143. Pietsch, W., 1976. Roll Pressing. Heyden & Son Ltd., London. Pietsch, W., 1991. Size Enlargement by Agglomeration. Wiley, England. Pietsch, W., 2002. Agglomeration Processes-Phenomena, Technologies, Equipment. Wiley-VCH Verlag GmbH, Weinheim. Schönert, K., Sander, U., 2002. Shear stresses and material slip in high pressure roller mills. Powder Technology 122, 136–144. Y.A. Yusof et al. / Chemical Engineering Science 60 (2005) 3919 – 3931 Sommer, K., Hauser, G., 2002. Flow and compression properties of feed solids for roll-type presses and extrusion presses. Powder Technology 130, 272–276. Thomson, F.M., 1984. Chapter 9 Storage of particulate solids. In: Fayed, M.E., Otten, L. (Eds.), Handbook of Powder Science and Technology. Van Nostrand Reinhold Company, New York. Tomas, J., Schubert, H., 1979. Particle characterisation. Partec 79, Nurnberg, Germany, pp. 301–319. 3931 Tüzün, U., 1987. Effects of consolidation and yield history on the measured angles of friction of particulate solids. In: Briscoe, B.J., Adams, M.J. (Eds.), Tribology in Particulate Technology. Adam Hilger, Bristol, pp. 38–62. Walker, D.M., 1966. An approximate theory for pressures and arching in hopper. Chemical Engineering Science 21, 975–997.