Pneumatic separation of hulls and meats from cracked soybeans

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copyfood and bioproducts processing 8 7 ( 2 0 0 9 ) 237–246

Contents lists available at ScienceDirect

Food and Bioproducts Processing

journa l homepage: www.e lsev ier .com/ locate / fbp

Pneumatic separation of hulls and meatsfrom cracked soybeans

Murilo D.M. Innocentinia,b,∗, Wellington S. Barizana,Maicon N.O. Alvesa, Reinaldo Pisani Jra,b

a Chemical Engineering Course, University of Ribeirão Preto-UNAERP, Av. Costábile Romano 2201,Ribeirânia, 14096-900 Ribeirão Preto, SP, Brazilb Post-graduation Program in Environmental Technology, University of Ribeirão Preto-UNAERP, Av. Costábile Romano 2201,Ribeirânia, 14096-900 Ribeirão Preto, SP, Brazil

a b s t r a c t

The dehulling process of cracked soybeans was experimentally investigated in this work. The mean Sauter diame-

ter (dv,s) of as-received material was 2.70 mm, with an average proportion of 95% meats and 5% hulls for a moisture

content of 11.8%. The true densities of hulls and meats were 1090 and 1267 kg/m3, with dv,s of 2.11 and 2.74 mm,

respectively. Hulls were mostly elutriated around 2.7–4.5 m/s and meats around 9.1–13.7 m/s. The overlap of ter-

minal velocity profiles required a combination of pneumatic and sieving operations for optimized separation. The

influence of particle concentration on continuous dehulling was investigated for three solid-to-air ratios (W/Q). The

procedure that maximized particle separation was a sequence of pneumatic dehulling with vs = 7.4–9.1 m/s and

W/Q = 1.05 kgsolids/m3air, followed by screening of lifted material with sieve ASTM no. 6 and a final pneumatic sepa-

ration of small hulls and meats at vs = 3.9 − 4.1 m/s. An industrial scale pneumatic dehuller was built and tested for

W = 6973 kg/h, vs = 7.6–8.2 m/s and W/Q = 0.97 kgsolids/m3air. The efficiency of the pneumatic device to remove hulls

from the cracked soybean was very high, with the recovery of meats with purity around 99%.

© 2008 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Soybean; Hulls; Pneumatic dehulling; Terminal velocity; Elutriation

1. Introduction

Soybean is the agricultural crop with the largest planted areain Brazil. The country was the second-largest world producerand exporter of this oilseed crop in 2007, just behind theUnited States. The expectation for the next years is still opti-mistic, especially because of the increasing demand of oilsources for the food industry and recently for the biodiesel pro-duction. This fact has motivated Brazilian industries to searchfor cost-reducing technologies to improve the market compet-itiveness of soybean byproducts, especially the meal fraction,a rich protein–carbohydrate–fiber ingredient for livestock andfood production.

Whole soybeans have a typical size of 5–8 mm (Duarte etal., 2004; Kocabiyik et al., 2004) and, on a dry weight basis, con-tain about 40% proteins, 20% lipids (oil), 35% carbohydrate andabout 5% ash. The grain comprises approximately 92% kernel

∗ Corresponding author at: Chemical Engineering Course, University of Ribeirão Preto-UNAERP, Av. Costábile Romano 2201, Ribeirânia,14096-900 Ribeirão Preto, SP, Brazil. Tel.: +55 16 3603 6784.

E-mail address: [email protected] (M.D.M. Innocentini).Received 19 July 2008; Received in revised form 31 October 2008; Accepted 5 November 2008

or meat (90% cotyledons and 2% germ) and 8% seed coat orhull (Erickson, 1995; Liu, 1997; Francis, 1999). Cleaned hullstypically have thickness of 0.4–0.5 mm (Raji and Famurewa,2005) and contain 9.4% crude protein and 86% complex car-bohydrates, a composition that makes hulls an importantsource of dietary fiber (Erickson, 1995). The insoluble car-bohydrate fraction of hull cell walls consists of 30% pectin,50% hemicellulose, and 20% cellulose (Gnanasambandam andProctor, 1999; Mullin and Xu, 2001). A unique aspect of soy-bean hulls is that the fiber content is low in lignin and ishighly digestible; therefore hulls are recognized as an excel-lent source of readily available energy in forage-based diets(Löest et al., 2001).

Soybean hulls are poor in oil content and are fairly abrasive.Thus it is convenient to remove them from the meat prior tothe oil extraction to avoid reduction in the oil yield and wearof flaking machines (Erickson, 1995).

0960-3085/$ – see front matter © 2008 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.doi:10.1016/j.fbp.2008.11.001

Author's personal copy238 food and bioproducts processing 8 7 ( 2 0 0 9 ) 237–246

Nomenclature

Ar Archimedes numberCD Drag coefficientcm mass concentration of particles in the air

stream (kg/m3)cv volumetric concentration of particles in the air

stream (m3/m3)D column diameter (m)d50 cut diameter of the cyclone collector (�m)di opening of the sieve i (mm)dP particle diameter (mm)dv,s average Sauter mass diameter (mm)g acceleration of gravity (m/s2)m number of sieve fractions in Eq. (1)n exponent in Eq. (5)P absolute ambient pressure (Pa)Q volumetric air-flow rate (m3/h)T temperature of the air-flow (◦C)vs superficial air velocity (m/s)vt terminal velocity of a single representative par-

ticle (m/s)vt,susp mean terminal velocity of particles in a suspen-

sion (m/s)wi mass fraction of particles retained in sieve i

(kg/kg)W mass flow rate of solids or Solids feeding rate

(kg/h)W/Q feed ratio or Solid-to-air ratio (kgsolids/m3

air)

Greek letters� fluid viscosity (Pa s)�f fluid density (kg/m3)�p true particle density (kg/m3)

In handling and processing of several agricultural prod-ucts, air is commonly used as a carrier for transport (Kılıckanand Güner, 2006) or for separating the desirable productsfrom unwanted materials (Gorial and O’Callaghan, 1990;Khoshtaghaza and Mehdizadeh, 2006). Therefore, the knowl-edge of terminal velocities of grain and impurity particlesbecomes essential for design and operation of pneumaticdevices.

Several equations for prediction of terminal velocitiesare available in the literature (Hawk et al., 1966; Kunii andLevenspiel, 1997; Fayed and Otten, 1997; El-Sayed et al.,2001) but their reliability for non-spherical particles is still acontroversial issue. For this reason, many researchers havemeasured the terminal velocities for various agricultural cropsand empirically related them to other physical features suchas size distribution, sphericity, porosity, moisture content, truedensity and bulk density (Dursun and Dursun, 2005; Polat etal., 2006; Cosckun et al., 2006; Gupta et al., 2007; Güner, 2007;Afonso Júnior et al., 2007; Zewdu, 2007).

Likewise, aerodynamic properties are important to designequipments for cleaning, drying, processing and storing soy-bean products (Soponronnarit et al., 2001; Polat et al., 2006).Duarte et al. (2004) found that whole soybean with mois-ture content of 10.90% (w.b.), equivalent sphere diameter of6.6 mm, true density of 1170 kg/m3 and sphericity of 0.897,have (has?) terminal velocity of approximately 14 m/s. Simi-lar investigations were carried out by Deshpande et al., 1993

and Polat et al. (2006). These latter authors found that forwhole soybean with geometric mean diameter of 5.62 mmand 6.7% moisture content, the values for sphericity, poros-ity, true density and terminal velocity were respectively 0.75,0.51, 1062.6 kg/m3 and 7.13 m/s. For 15.3% moisture content,these values changed respectively to 0.72, 0.442, 1086.4 kg/m3

and 9.24 m/s.The systems used to remove hulls in industrial soy-

bean plants today vary considerably in design and capacity.However, they all share the same principle when the basicoperation is examined: the difference in terminal velocitiesof hulls and meats caused by their very different morpholo-gies: the hulls being flatter, the meats more rounded. If theequipment is properly sized so that the air velocity is below theterminal velocity of the meats but above that of the hulls, then,theoretically, only hulls are aspirated and a good separation isachieved.

There are a number of air separators available, and usu-ally the more simple and non-adjustable are the most suitablein terms of capacity and maintenance (Boling, 1979). One ofthe simplest pneumatic hull separators is a vertical cylindricalcolumn in which previously cracked soybeans are fed down-wardly through a rising stream of air. The lighter particles,comprised of hulls and fines, are aspirated out of the top of thecolumn while the heavier meats continue to cascade down-ward. The aspirated hulls are collected in a cyclone, passed toa toaster to destroy urease activity, ground to the desired par-ticle size and either pelleted or sold as bulk (Erickson, 1995;Wrigley, 2004).

Typical industrial recommendation is that soybeans shouldbe cracked into a few pieces only, in order to facilitate therelease of hulls from meats but also to prevent the genera-tion of fines. In practice, however, size distribution may bebroad, since the fineness of the material depends on featuresof the particles, such as moisture content and temperature,and of the cracking machine, such as the rollers spacing androtating speed. A typical problem found in industry is thefact that as hull and meat particles are progressively crackedin smaller sizes, their terminal velocities become closer, pre-cluding efficient separation (Boling, 1979). The general rule isthat the generation of coarser crushed material enables easierhull aspiration without taking away too much oil stock withthe fines. If a broad particle size distribution is generated inthe cracking mill, then mechanical separation by sieving isadvised ahead of air sifting to remove fines. So, dehulling sys-tems based on sequences of sieving and aspiration steps arestill common in the industry.

Unfortunately, much of the aerodynamic data required tooptimize the dehuller operation has been gathered by trialand error methods in the soybean industry, especially becauseoptimum separation is ultimately related to the size distribu-tions of hull and meat particles, which in turn may vary in eachindustry according to the crushing operation or other physi-cal and mechanical features of the grains. Data concerningthe ranges of terminal velocities of soybean fractions are stillscarce in the literature. Boling (1979) presented a very goodqualitative discussion about principles of soybean dehullingand strategies of efficient hull removal by combination ofscreening and aspiration, but no useful data was reported.Physical properties of soybean meal, including particle sizeand particle size distributions, bulk density and true density,and repose angle and drain angle, were studied in the labora-tory by Wang et al. (1995), but no terminal velocity data wasreported.

Author's personal copyfood and bioproducts processing 8 7 ( 2 0 0 9 ) 237–246 239

Fig. 1 – Schematic illustration of apparatus fordetermination of terminal velocities of fractions of crackedsoybeans.

From the context previously described, it becomes clearthat efficient operation of pneumatic dehullers depends onthe data of both particle size distributions of hulls and meatsand of their terminal velocities. In this work, an experimentalwork is conducted to collect and correlate such data in orderto find operational conditions to optimize soybean dehulling.

2. Materials and methods

Cracked soybeans were supplied by a Brazilian soybean pro-cessing industry and extracted directly from its productionline. Particle size distribution of as-received material wasobtained by vibrating screening with ASTM standard sievesnos. 3.5, 4, 6, 8, 10, 18, 35 and 50, corresponding to openings of5.60, 4.75, 3.35, 2.36, 2.00, 1.00, 0.50 and 0.30 mm, respectively.Tests were carried out in triplicate with a mass of 100 g foreach trial.

Hulls and meats were then manually separated from sam-ples of the original material and had their true densitiesevaluated by helium pycnometry (AccuPyc 1330). Particle sizedistributions of both fractions were obtained by screening withthe same previously mentioned sieves.

Terminal velocities of cracked soybean fractions were mea-sured in the pneumatic apparatus schematically shown inFig. 1. Air-flow was supplied by a 4.5 hp centrifugal blowerand insufflated upward through a vertical transparent col-umn with an inside diameter of 48 mm and a height of 1.5 m.A high-efficiency Stairmand-type cyclone was dimensionedaccording to Iozia and Leith model (Coury et al., 2004) andbuilt to collect the particles elutriated by the air-flow. The cutparticle size for 50% collection efficiency (d50) was 1.9 �m foroperation at ambient conditions (T = 30 ◦C and P = 93.3 kPa) andair-flow rate of Q = 60 m3/h.

A sample of 50 g was placed on an 80-mesh screen basisfixed at the bottom of the column. A digital rotating vaneanemometer (Minipa MDA-11) measured the velocity at the airoutlet of the cyclone and values were corrected for the velocity

based on the cross-sectional area of the free column (superfi-cial air velocity, vs). Air-flow temperature was kept at 26–30 ◦Cand relative air humidity was not controlled during the tests.For each superficial air velocity increased stepwise from 0 to18 m/s, a fraction of particles (hulls and meats) was fluidizedand then elutriated by the air-flow to the cyclone. After afew minutes, when no conveying was observed in the tube,the mass of particles collected in the bottom of the cyclonewas removed, weighed and manually classified as meat orhull fractions. A higher air velocity was then set and the testrepeated until the complete conveying of the sample fromthe column. The theoretical cyclone collection efficiency washigher than 99% for particles larger than 55 �m (equivalentto ASTM sieve no. 270) in all velocity range tested. Based onthe experimental size distribution of cracked soybeans, it wasassumed that no loss occurred and the mass collected in thecyclone represented the whole particle mass pneumaticallyconveyed from the column for each superficial velocity tested.

The influence of the meat particle size on the terminalvelocity was also investigated. A 50 g sample of cracked soy-bean was manually classified to eliminate the whole hullcontent, and a new pneumatic test was carried out only withthe meat portion (∼47 g). For each superficial velocity, the frac-tion of meats elutriated was weighed and classified by sieving.The average Sauter mass diameter (dv,s) of each fraction wascalculated by Eq. (1) according to Kunii and Levenspiel (1991):

dv,s = 1∑m

i=1wi/di

(1)

in which wi is the mass fraction of particles retained in eachaverage sieve size di and m is the number of fractions obtainedby sieving.

Terminal velocity vt for each meat fraction was predictedaccording to Perry et al. (2008) as

vt =(

43

gdp(�p − �f)�fCD

)0.5

(2)

in which the drag coefficient (CD) is found in terms of theArchimedes number (Ar) as

CD = 432Ar

(1 + 0.0470Ar2/3) + 0.5171 + 154Ar−1/3

(3)

Ar = 43

(�p − �f)g�fdp3

�2(4)

in which �p and �f are respectively the particle and fluid den-sities, g is the acceleration of gravity, � is the fluid density,and dp is mean the particle diameter (used here as the averageSauter diameter obtained by Eq. (1) for meat fractions collectedin each sieve).

Eq. (3) is valid for rising light spheres with Ar in the range0–13,000. For Ar > 13,000, CD = 0.95. It is worth noting that fornon-spherical particles the actual terminal velocity dependsupon their shape and orientation with respect to the directionof motion, which is a function of Reynolds number (Perry et al.,2008). In this work, however, no experimental shape factor wasdetermined for meats and the particle size was assumed asthe diameter of the sphere passing the same minimum squareaperture, i.e., the sieve diameter (dp = di).

Based on the batch test results, a suitable superficial veloc-ity range to maximize the separation of hulls and meats was


selected and a new set of experiments was carried out. Thepneumatic apparatus was adapted for continuous downwardfeed of cracked soybean particles, with the upward air-flowprovided by air suction from the atmosphere at the bottom ofthe separation column. Steel rods of 4 mm diameter were fixedtransversally in 10 sections of the column in order to facilitatethe dispersion of the falling particulate phase into the air-flow.Reduction of free flow area in these sections was about 18.1%,with a proportional increase in air velocity. The particles withterminal velocity below that of the air were carried upwardby the air stream and those with terminal velocity above thatof the air fell downwardly through the column. Three ratiosbetween the feeding rate of solids (W) and the air intake tothe column (Q) were tested. Superficial air velocity and tem-perature were constant during the three tests. Hull and meatcontents in both rising and falling streams were evaluated byvisual inspection of collected samples. The amount of parti-cles elutriated to the cyclone was screened and the fractionpassed through the sieve was subjected to a secondary pneu-matic separation.

Finally, based on the information collected in the labora-tory tests, an industrial scale dehuller was designed, built andtested for validation of results. A high throughput swift-typecyclone was designed (Coury et al., 2004) and built for collec-tion of aspirated hull particles. Samples were collected at thebasis of the column for determination of separation efficiency.

3. Results and discussion

3.1. Composition and particle size distribution ofcracked soybeans

Analysis of as-received cracked soybean samples indicated aweight content of 95.06% meats and 4.94% hulls. Originalmoisture content of the raw material was 11% (dry basis).These values were confirmed by triplicates (standard devia-tion of 13%) and revealed a hull content lower than the valueof 8% cited by the technical literature (Brooker, 1992; Liu, 1997;Sessa and Wolf, 2001), but very close to the expected indus-trial yield of 5%. The discrepancy may be due to the presenceof some hulls still stuck to larger meat pieces and the diffi-culty to visually distinguish hull and meat particles in the finerfractions.

The true densities (�p) of hulls and kernels were respec-tively 1090 and 1267 kg/m3, values that are in the same range ofthose reported by Deshpande et al. (1993) for soybean kernels(1216–1124 kg/m3), by Duarte et al. (2004) for whole soybeans(1170 kg/m3) or by Polat et al. (2006) also for whole grains(1062–1086 kg/m3).

Data concerning particle size distributions of cracked soy-bean and their fractions are presented in Figs. 2–4. Fig. 2 showsthe size distribution of as-received cracked grains, with noseparation between hulls and meats. A broad distribution isobserved, probably resulting from an aggressive breakage pro-cess in the mill. The mean Sauter diameter (dv,s) of particlescalculated from Eq. (1) was 2.70 mm.

The material collected in each sieve was then manuallyseparated in hull and meat fractions and the relative contentis shown in Fig. 3. Due to the difficulty in visual inspection,the fraction of fines collected in sieve no. 50 was considered ashulls particles only. Hulls and meats were broadly distributedamong all other sieves. For instance, material collected in sieveno. 4 is composed of around 88% of meats and 12% of hulls,

Fig. 2 – Discrete and cumulative particle size distributionsof as-received cracked soybean obtained by sieving. (#) Theaverage particle size is the arithmetic average between asieve opening di and the next larger opening di+1.

Fig. 3 – Proportions of hulls and meats collected in eachsieve. (#) The average particle size is the arithmetic averagebetween a sieve opening di and the next larger opening di+1.

while content in sieve no. 18 is 91.5% of meats and 8.5% ofhulls. This aspect indicates that an efficient hull separationbased only on sieving is unfeasible in the present case.

Fig. 4 shows the individual size distributions for hulls andmeats. The similarity of profiles indicates that the crackingprocess affected both fractions in the same way. Breakage ofhulls was only slightly more aggressive, as about 11% of theirparticles were retained in sieves nos. 35 and 50 (openings of0.75 and 0.30 mm). Most of material was retained in sieves nos.6 and 8, and the average Sauter diameters (dv,s) for hull andmeat particles were respectively 2.11 and 2.74 mm.

Fig. 4 – Particle size distributions of hulls and meatsobtained by sieving of cracked soybeans.


Fig. 5 – Details of meats and hulls separated by sieving andby pneumatic procedures (scale in centimeters). (a) Materialretained in the ASTM sieve no. 6; (b) Material elutriated by3.4 m/s air-flow.

Fig. 6 – Proportion of hulls and meats lifted at eachsuperficial air velocity.

3.2. Settling properties of the cracked soybeans

Results of aerodynamic classification tests are shown inFigs. 5–8. It becomes clear here that hull and meat fractions ofsame sieve sizes have in fact very different terminal velocityranges. Hull particles are less dense than meats and also have

Fig. 7 – Mass distributions of hulls and meats lifted at eachsuperficial air velocity.

Fig. 8 – Cumulative mass fractions of hulls and meats liftedup to each superficial air velocity.

a skeletal hemispherical shape. Such combination of featuresimplies in a high specific drag area (surface area in the planperpendicular to the flow direction) and explains why hullsare preferentially elutriated at lower superficial air velocities.Fig. 5 shows pictures illustrating that large hulls retained insieve no. 6 (di = 3.35 mm) have in fact terminal velocity com-parable with that of small meat particles passing throughsieve no. 18 (di = 1.50 mm). The consequence is that fractionsof cracked soybeans elutriated up to 4.5 m/s are basically com-posed of hulls and fractions above 6.8 m/s are mostly meats(Fig. 6).

Individual profiles of lifted hull and meat particles areshown in Fig. 7. Hulls are mostly lifted in a narrow velocityrange (2.7–4.5 m/s), while elutriation of meats is more pro-nounced from 9.1 to 13.7 m/s. Fig. 8 shows the triplicate ofair-flow separation tests, confirming the very different velocityrange for lifting of hulls and meats.

The overlap of lifting profiles around 6–8 m/s shown inFigs. 7–8 indicates that the separation of meats and hulls intotwo pure streams is troublesome if just one single pneumaticoperation is intended. It is worth noting that such overlap infact depends on the particle size distribution produced in thecracking mill and it helps to explain why commercial dehullersoperating at same air velocities may have different separationperformances from industry to industry. A useful procedure inthis case is to operate at an air velocity high enough to mini-mize the presence of hulls in the downward stream of meats.In this situation, the small meats may also be lifted, but theycan be subsequently separated from large hulls and recoveredby screening.

The full meat content of a cracked soybean sample wasmanually separated from hulls and the meat fraction was sub-jected to a new pneumatic test to determine the relationshipbetween terminal velocity vt and average particle Sauter diam-eter dv,s. Curves in Fig. 9 show a relatively good agreement forexperimental and predicted values of terminal velocity, espe-cially if considering that no shape factor was applied for sizecorrection of non-spherical particles in Eqs. (2–4). Maximumdeviation was 18% for the largest particles.

Information in Fig. 9 also helps to choose a suitable set ofsieves that provides good size separation of meat and hullsdragged together to the cyclone. For instance, for air veloc-ity in the range of 7–8 m/s, the average Sauter diameter (dv,s)of lifted meats is around of 1.5 mm. These particles can beremoved from larger hulls afterwards by screening with ASTMsieves nos. 12–14 (openings between 1.4 and 1.7 mm). Another


Fig. 9 – Terminal velocities for meat particles of differentaverage sizes.

pneumatic separation can then provide the final separation ofsmall hulls and meats that pass though these sieves.

3.3. Continuous pneumatic separation system

An important aspect for continuous pneumatic dehulling isthat terminal velocities are affected not only by the featuresof particles (shape, texture, size, moisture and density) andfluid (density and viscosity) but also by the particle concen-tration inside the column and by wall effects. It is well knownfrom the literature that the terminal velocity of a single parti-cle is significantly reduced by the presence of other particles(Fayed and Otten, 1997; Klinzing et al., 1997; Perry et al., 2008).The fluid is forced to percolate through the void spaces notoccupied by particles. Since the effective flow area in the sec-tion containing particles is smaller than the cross-sectionalarea of the tube, the interstitial velocity becomes higher thanthe superficial velocity, based on the free tube section. Con-sequently, there is an increase in the air resistance force thathinders the particle settling and reduces its terminal veloc-ity. According to Fayed and Otten (1997), the effect of particleconcentration on the terminal velocity of suspensions can be

predicted by Richardson and Zaki equation:

vt,susp = vt(1 − cv)n (5)

in which vt,susp is the mean terminal velocity of particles in asuspension, vt is the terminal velocity of a single representa-tive particle, cv is the volumetric concentration of particles inthe suspension and n is an exponent that depends on the ratiobetween the particle size (dp) and the column diameter (D) andon the Reynolds number. Typically a value of n = 4.7 is accept-able for a large number of real suspensions. The volumetricconcentration cv can be determined by

cv = cm

�p(6)

in which cm is the mass concentration of particles (for exam-ple, kgsolids/m3

air) in the suspension and �p is the particledensity (kgsolids/m3

solids).Eq. (5) predicts that a reduction of 11% in terminal veloc-

ity is expected for a suspension with 2% of solids in volume;reduction may reach up to 50% for 10% suspensions. A simi-lar effect on terminal velocity is observed when the particlesize becomes comparable to the size of the column cross-section. According to Duarte et al. (2004), the terminal velocityof a single soybean with diameter of 6.6 mm, true densityof 1170 kg/m3 and moisture of 10.9% is reduced from 14.7 to2.7 m/s when the vertical column diameter (D) is reduced from60 to 20 mm.

For optimization of pneumatic dehullers, it is importantto know how the separation velocity of hulls and meats isaffected by the particle concentration inside the aspirationcolumn. One parameter useful for analysis is the ratio betweenthe feeding rate of solids (W) and the volumetric air-flow rate(Q). For practical purposes, the W/Q ratio can be taken as ameasure of the mass concentration cm.

For a fixed volumetric flow rate Q, the increase in the solidsfeeding rate W imply in larger processing capacity for theplant, but then solids concentration becomes higher inside the

Fig. 10 – Three-step procedure used for separation of hulls and meats. Numbers 1–7 refer to the streams for which analysisis shown in Tables 1 and 2.


Table 1 – Operational conditions for pneumatic separation at three different solids/air ratios (stream labels are as definedin Fig. 10).

Operational conditions for pneumatic separation Test 1 Test 2 Test 31.05a 2.13a 2.98a

Stream 1Air-flow rate, Q (m3

air/h) 48.2 48.2 48.2Solids feeding, W (kgsolids/h) 50.6 102.7 143.6Air velocity, vs (free section, m/s) 7.4 7.4 7.4Air velocity, vs (sections with rods, m/s) 9.1 9.1 9.1Volumetric concentration of particulate phase in the column, cv (%) 0.08 0.17 0.24

0.18a 0.21a 0.19a

Stream 5Air-flow rate, Q (m3

air/h) 25.2 25.2 25.2Solids feeding, W (kgsolids/h) 4.43 5.32 4.87Air velocity, vs (free section, m/s) 3.9 3.9 3.9Air velocity, vs (sections with rods, m/s) 4.7 4.7 4.7Volumetric concentration of particulate phase in the column, cv (%) 0.014 0.017 0.015

a W/Q ratio (kgsolids/m3air).

column, which decreases the terminal velocities and worsensthe dispersion of the particulate phase. Besides, pressure dropthrough the column becomes higher, as well as the power con-sumption. On the other hand, the increase of the air intake Qfor a fixed solids feeding W provides a better particle disper-sion that enhances the separation efficiency, but the blowerpower consumption and the size of downstream equipmentsare proportionally affected. In fact, the more attractive W/Qratios depend on an overall analysis of the process and exper-imental investigation is required to evaluate how separationefficiency is affected.

In this work, a three-step strategy was adopted for evalua-tion of the influence of the W/Q ratio for continuous dehulling.Fig. 10 illustrates the separation sequence with the location ofall streams analyzed. The three steps were:

Step 1: Pneumatic separation at continuous scale of crackedsoybeans with a superficial air velocity high enough to min-imize the hulls fraction in the downward stream of meats,regardless of the possibility of conveying part of meat parti-cles to the cyclone. Three W/Q ratios were tested: 1.05, 2.13 and2.98 kgsolids/m3

air. Based on the batch tests results (Figs. 6–8),the superficial air velocity chosen was vs = 7.4 m/s, with air-flow rate fixed at Qair = 48.2 m3/h in all tests. The actual airvelocity in the sections containing the transversal rods to facil-itate dispersion was higher (9.1 m/s).

Step 2: Sieving of solids collected in the cyclone during step1. Since larger hulls and small meats have similar terminalvelocities, they are expected to be conveyed together to thecyclone. In this case, their separation can be carried out bysieving. Based on particle distributions shown in Figs. 3 and 4,a single ASTM sieve no. 8 (di = 2.36 mm) was chosen. Particlesretained in this sieve are expected to be only large hull pieces.The fraction collected in the pan below the sieve is expectedto be a mixture of small meat and hull particles.

Step 3: Pneumatic separation at continuous scale of smallmeats and hulls in the fraction that passed through thesieve no. 8 in Step 2. Based on the separation profile inFig. 6, the chosen superficial air velocity was vs = 3.9 m/s(4.7 m/s in the sections with rods). Therefore, air-flow ratewas set at Qair = 25.2 m3/h. The W/Q ratio was kept fixed at≈0.20 kgsolids/m3

air, lower than the ratios employed in step 1.This happened as the velocity range for separation of small

hulls and meats is narrower and efficient particle dispersionis required to prevent settling of hulls or drag of meats.

Three tests were carried out using the procedure previouslydescribed and schematized in Fig. 10. Table 1 gives a summaryof operational conditions for the pneumatic separations (Steps1 and 3).

Samples of particles in all streams identified in Fig. 10 wereweighed and visually inspected for determination of hull andmeat contents. Table 2 provides the proportions of hulls andmeats found in each stream, data that are helpful to deter-mine the efficiency of purification obtained by pneumatic andsieving operations for the different W/Q ratios.

The first pneumatic separation was efficient to removehulls from the fraction collected at the bottom of the aspi-ration column (stream 2). The purity of this meat stream wasvery high and only slightly reduced (from 0.07% to 0.27% hullcontent) by the increase of the W/Q ratio. Similarly, the pro-portion of meats in the stream lifted to the cyclone (stream3) was kept fairly constant (57–61%). However, there wasa significant increase in the size of elutriated meats, con-firmed by the increase of their fraction retained in sieve no.8 (stream 4). For operation at a W/Q = 1.05 kgsolids/m3

air, only6.90% meats were collected by screening, but this percentageincreased five times (34.38%) for a threefold increase in theratio to W/Q = 2.98 kgsolids/m3

air. It is not favorable to the pro-cess, because meat particles present in stream 4 are no longerrecovered.

Based on an average particle density (�p) of 1258 kg/m3

for the cracked soybean and ignoring the particle holdupinside the separation column, the tested ratios of 1.05, 2.13and 2.98 kgsolids/m3

air corresponded respectively to very dilutevolumetric concentrations (cv) of 0.08%, 0.17% and 0.24%,calculated from Eq. (5). According to Eq. (6), reduction in ter-minal velocity is lower than 1.4% for the highest W/Q ratio of2.98 kgsolids/m3

air. These values are too low to explain the lift-ing of large meats to the cyclone. A more logical explanationfor the observed behavior is the inhomogeneous dispersion ofparticles, especially near the feeding zone located at the upperposition of the column. Occasional disturbances in the flowof falling particles may have caused intermittent increase ofinterstitial air velocity and upward acceleration of meat parti-cles toward the linking pipe to the cyclone, where air velocity is


Table 2 – Relative contents of meats and hulls in each stream for tests at three different solids/air ratios (stream labels areas defined in Fig. 10).

Mass treated and contents of meats and hulls in each stream W/Q ratio (kgsolids/m3air)

1.05 2.13 2.98

Stream 1—fraction fed into the separation columnMass passing through the stream, M1 (g) 997.5 998.4 997.3Meats (%) 95.69 95.22 95.47Hulls (%) 4.31 4.78 4.53

Stream 2—fraction collected at the bottom of the columnMass passing through the stream, M2 (g) 897.6 888.4 887.5Meats (%) 99.98 99.93 99.73Hulls (%) 0.02 0.07 0.27

Stream 3—fraction lifted to the cycloneMass passing through the stream, M3 (g) 99.8 110.1 109.8Meats (%) 57.11 57.21 61.04Hulls (%) 42.89 42.79 38.96

Stream 4—fraction retained in sieve no. 8Mass passing through the stream, M4 (g) 26.1 27.3 38.1Meats (%) 6.90 14.91 34.38Hulls (%) 93.10 85.09 65.62

Stream 5—fraction fed into the separation column (W/Q ratio ≈0.20 kgsolids/m3air)

Mass passing through the stream, M5 (g) 73.8 82.8 71.7Meats (%) 74.86 71.16 75.21Hulls (%) 25.14 28.84 24.79

Stream 6—fraction collected at the bottom of the columnMass passing through the stream, M6 (g) 18.5 23.9 17.8Meats (%) 100.00 100.00 100.00Hulls (%) 0.00 0.00 0.00

Stream 7—fraction lifted to the cycloneMass passing through the stream, M7 (g) 55.2 58.9 53.9Meats (%) 0.00 0.00 0.00Hulls (%) 100.00 100.00 100.00

higher (>20 m/s) and capable of conveying even the largest par-ticles. A helpful procedure to minimize this effect is to increasethe column height above the feeder, so that large dense par-ticles occasionally lifted can be decelerated and settled againbefore being captured by the linking pipe.

The second pneumatic separation (Step 3) was also suc-cessful to treat the mass of fine meats and hulls that passedthrough sieve no. 8 (stream 5). Qualitative inspection of sam-ples from streams 6 and 7 indicated that the fraction settledto the bottom of the column was basically composed of meatswhile the fraction conveyed and collected in the cyclone wascomposed of fine hull particles and powder. Due to the smallparticle size, it was difficult to conduct a systematic man-

ual separation of fines into hulls and meats. For this reason,stream 6 was considered 100% meats and stream 7 was 100%hulls. This assumption is fairly acceptable, since the mass bal-ance including all streams indicated the same overall contentas previously obtained in the batch tests (∼95% meats and∼5%).

Based on the mass balance and on the content analysisfor each stream given in Table 2, the yield and the purity ofmeats and hulls streams resulted from the three-step separa-tion process was determined, as summarized in Fig. 11. Theyield of meats was based on the sum of mass of all particlescollected in streams 2 and 6, while the yield of hulls was basedon streams 3 and 6. Data in Fig. 11 confirm that in terms of

Fig. 11 – Summary of yields and purities of hulls and meats obtained for each W/Q ratio.


Fig. 12 – Details of pneumatic soybean dehuller built for pilot scale tests: (a) aspiration column (D = 0.580 m, H = 5 m); (b)rotary feeder (W = 6973 kg/h); (c) centrifugal blower (Q = 7206 m3/h); (d) swift-type cyclone; (e) transversal rods inside columnto help particle dispersion.

yield or purity of meats recovered from the process, all W/Qratios were efficient. However, the loss of meats present inthe hull streams became considerable for W/Q ratios above1.05 kgsolids/m3

air.

3.4. Tests in the pilot plant

Fig. 12 shows some pictures of the pneumatic soybeandehuller of industrial scale that was built and used for valida-tion of the laboratory results. The vertical column had insidediameter of 0.580 m and total height of about 5 m. The usefulheight for particle separation below the feeder was 2 m, withfour sections of 3/4 in. metallic rods fixed transversally at each50 cm to facilitate dispersion of particles. The open flow areain sections with rods was reduced in about 8%. The columnheight above the feeder was about 2.5 m in order to minimizeentrance disturbances verified in the laboratory tests. Thesolids feeding rate was W = 6973 kg/h, which is fairly compati-ble with typical scale of commercial dehullers in the soybeanindustry. The volumetric air-flow rate was set at Q = 7206 m3/hin order to provide a W/Q ratio of 0.97 kgsolids/m3

air, a valuevery close to the best found in the laboratory tests. Air veloc-ity through the column was 7.6 m/s (8.2 m/s in the sectionswith rods). From the 230.5 kg treated in the pneumatic devicein 119 s, 203.3 kg (88.2%) were collected at the bottom of thecolumn and 27.2 kg (11.8%) were lifted and recovered at thecyclone. Screening analysis revealed that 18.49% of the con-veyed fraction was retained in sieve no. 6 (only large hulls),31.08% in sieve no. 9, 42.20% in sieve no. 16 and 8.23% at thebottom pan (only fines). Visual inspection and weight analysisconfirmed the efficiency of the pneumatic device, as the purityof the meat stream at the bottom of the dehuller was around99%, while only small meats were the aspirated out togetherwith hulls to the cyclone.

4. Conclusions

In a wider context, results in this investigation are useful toplan a suitable procedure for continuous dehulling of agri-cultural crops, especially for grains that present considerabledifferences in density and morphology between their hull andmeat fractions. For cracked soybean, it was observed that thekey feature that provides an efficient pneumatic dehulling isthe terminal velocity ranges of hulls and meats. Broad par-ticle size distributions tend to cause the overlap in terminalvelocities of both fractions that precludes good separationand makes necessary the post-treatment of lifted particles byscreening.

With respect to the tests and data collected in this work,the following specific conclusions could be drawn:

(1) Particle size distribution of as-received cracked soybeanwas relatively broad, probably resulting from an aggressivebreakage in the milling process. The mean Sauter diame-ter (dv,s) of particles was 2.70 mm. Visual inspection andweight analysis determined an average proportion of 95%meats and 5% hulls for a moisture content of 11.8%.

(2) The true densities of hulls and meats were respectively1090 and 1267 kg/m3, values that are in agreement withthose reported in the literature. Screening analysis showedsimilar particle size distributions for hulls and meats, withaverage Sauter diameters of 2.11 and 2.74 mm, respec-tively.

(3) Hull and meat fractions of same sieve sizes had differ-ent terminal velocity ranges, caused by the differences indensity and shape. Hulls were mostly elutriated in a nar-row velocity range (2.7–4.5 m/s), while lifting of meats wasmore pronounced from 9.1 to 13.7 m/s.


(4) The overlap of lifting profiles around 6–8 m/s and the sim-ilarity of particle size distributions of hulls and meatsindicated that their optimized separation must be basedin a combination of pneumatic and sieving operations.

(5) A three-step procedure for continuous separation of hullsand meats was evaluated. The chosen air velocity to pro-duce a purified stream of meats was in the range of7–9 m/s. The influence of particle concentration duringcontinuous pneumatic dehulling was investigated in lab-oratory for three solid-to-air ratios (W/Q).

(6) The increase in the W/Q ratio did not affect the purityand the yield of meat streams, but implied in an increasein the lifting and carriage of large meat particles tohull streams, with consequent decrease of their purity.The procedure that maximized particle separation was asequence of pneumatic dehulling with vs = 7.4–9.1 m/s,W/Q = 1.05 kgsolids/m3

air, followed by screening of mate-rial dragged to the cyclone with sieve ASTM no. 6 and afinal pneumatic separation of small hulls and meats thatpassed through sieve at vs = 3.9–4.1 m/s.

(7) A pilot scale pneumatic dehuller was built and testedfor a feeding rate of 6973 kg/h, vs = 7.6–8.2 m/s andW/Q = 0.97 kgsolids/m3

air. The efficiency of the pneumaticdevice to remove hulls from the cracked soybean wasvery high, with the recovery of meats with purity around99%.

Acknowledgments

The authors wish to acknowledge UNAERP for the financialsupport and Masiero Industrial S.A. and Produtos AlimentíciosOrlândia S.A. for the supplies of cracked soybeans and also forthe technical discussions.

References

Afonso Júnior, P.C., Corrêa, P.C., Pinto, F.A.C. and Queiroz, D.M.,2007, Aerodynamic properties of coffee cherries and beans.Biosystems Engineering, 98: 39–46.

Boling, F., 1979, Principles of hull removal from oilseed. Oil MillGazetteer, February: 29–31.

Brooker, D.B., (1992). Drying and Storage of Grains and Oilseeds (1stedition). (Springer).

Cosckun, B.M., Yalcin, I. and Özarslan, C., 2006, Physicalproperties of sweet corn seed (Zea mays saccharata Sturt.).Journal of Food Engineering, 74: 523–528.

Coury, J.R., Pisani, R., Jr. and Hung, Y.T., 2004, Cyclones, Handbookof Environmental Engineering (2nd edition, vol. 1, pp. 97–152).

Deshpande, S.D., Bal, S. and Ojha, T.P., 1993, Physical properties ofsoybean. Journal of Agricultural Engineering Research, 56(2):89–98.

Duarte, M.E.M., Mata, M.C., Torres, H.L.H. and Silveira, V., Jr., 2004,The shape effect of the fall tunnel on aerodynamic parametersof soy grain, In Proceedings of the 14th International DryingSymposium (IDS 2004), vol. B São Paulo, Brazil, , pp. 1496–1501.

Dursun, E. and Dursun, I., 2005, Some physical properties ofcaper seed. Biosystems Engineering, 92(2): 237–245.

El-Sayed, A.S., Yahaya, R., Wacker, P. and Kutzbach, H.D., 2001,Characteristic attributes of the peanut (Arachis hypogaea L.) forits separation. International Agrophysics, 15: 225–230.

Erickson, D.R. (ed) 1995, Practical Handbook of Soybean Processingand Utilization.

Fayed, M. and Otten, L., (1997). Handbook of Powder Science andTechnology (2nd edition). (Chapman & Hall, New York), p. 864

Francis, F.J., (1999). Encyclopedia of Food Science and Technology (2ndedition). (John Wiley & Sons).

Gnanasambandam, R. and Proctor, A., 1999, Preparation of soyhull pectin Food Chemistry, 65:461–467.

Gorial, B.Y. and O’Callaghan, J.R., 1990, Aerodynamic properties ofgrain/straw materials. Journal of Agricultural EngineeringResearch, 46: 275–290.

Güner, M., 2007, Pneumatic conveying characteristics of someagricultural seeds. Journal of Food Engineering, 80: 904–913.

Gupta, R.K., Arora, G. and Sharma, R., 2007, Aerodynamicproperties of sunflower seed (Helianthus annuus L.). Journal ofFood Engineering, 79: 899–904.

Hawk, A.L., Broker, D.B. and Cassidy, J.J., 1966, Aerodynamiccharacteristics of selected farm grains. Transactions of theASAE, St. Joseph, Michigan, 9(1): 48–51.

Khoshtaghaza, M. and Mehdizadeh, R., 2006, Aerodynamicproperties of wheat kernel and straw materials. AgriculturalEngineering International: the CIGR Ejournal. Manuscript FP05 007, vol. VIII.

Kılıckan, A. and Güner, M., 2006, Pneumatic conveyingcharacteristics of cotton seeds. Biosystems Engineering, 95(4):537–546.

Klinzing, G.E., Marcus, R.D., Rizk, F. and Leung, L.S., (1997).Pneumatic Conveying of Solids: A Theoretical and PracticalApproach (2nd edition). (Chapman & Hall (Powder TechnologySeries), London, UK).

Kocabiyik, H., Akta, T. and Kaysoglu, B., 2004, Porosity rate ofsome kernel crops. Journal of Agronomy, 3(2): 76–80.

Kunii, D. and Levenspiel, O., (1991). Fluidization Engineering (2ndedition). (Butterworth-Heinemann, Newton, MA, USA).

Kunii, D. and Levenspiel, O., 1997, Circulating fluidized-bedreactors. Chemical Engineering Science, 52(15): 2471–2482.

Liu, K., (1997). Soybeans: Chemistry, Technology and Utilization.(Chapman and Hall, New York), p. 532

Löest, C.A., Titgemeyer, E.C., Drouillard, J.S., Blasi, D.A. andBindel, D.J., 2001, Soybean hulls as a primary ingredient inforage-free diets for limit-fed growing cattle. Journal ofAnimal Science, 79: 766–774.

Mullin, W.J. and Xu, W., 2001, Study of soybean seed coatcomponents and their relationship to water absorption.Journal of Agricultural and Food Chemistry, 49: 5331–5335.

Perry, R.H., Benskow, L.R., Beimesch, W.E., et al., (2008). Perry’sChemical Engineers’ Handbook (8th edition). (McGraw-Hill, NewYork).

Polat, R., Atay, U. and Saglam, C., 2006, Some physical andaerodynamic properties of soybean. Journal of Agronomy, 5(1):74–78.

Raji, A.G. and Famurewa, J.A.V., 2005, Parameters affecting millingqualities of undefatted soybeans (Glycine max, L. Merrill) (2):selected physical properties. International Journal of FoodEngineering, 1 (5). Available athttp://www.bepress.com/ijfe/vol1/iss5/art6/. Access on20/01/2008.

Sessa, D.J. and Wolf, W.J., 2001, Bowman-Birk inhibitors insoybean seed coats. Industrial Crops and Products, 14(1):73–83.

Soponronnarit, S., Swasdisevi, T., Wetchacama, S. andWutiwiwatchai, W., 2001, Fluidised bed drying of soybeans.Journal of Stored Products Research, 37: 133–151.

Wang, Y., Chung, D.S. and Spillman, C.K., 1995, Physicalproperties of soybean meal. Cereal Chemistry, 72(6): 523–526.

Wrigley, C. (Ed.). 2004, Encyclopedia of Grain Science. (AcademicPress)

Zewdu, A.D., 2007, Aerodynamic properties of tef grain and strawmaterial. Biosystems Engineering, 98: 304–309.

Pneumatic separation of hulls and meats from cracked soybeans

Documents

Transcript of Pneumatic separation of hulls and meats from cracked soybeans