The development of a horizontal elutriator: the infrasizer MK III

THE DEVELOPMENT OF A HORIZONTAL ELUTRIATOR: THE INFRASIZER r~K II I

TECHNISCHE HOGESCHOOL OElFT

LUCHTVAART -HJ

RIJIMT

E\IAARH[CIf~

~IE:<

by

BIBLQ"rHE~ • Kluyverweg 1 - DELFT

S. Raimondo, A. A. Haasz and B. Etkin

Mareh, 1979

05

JUKI '979

UTIAS Report No.

235

eN

ISSN

0082-5255

f

THE DEVELOPMENr OF A HORIZONrAL ELUTRIATOR: THE INFRASlZER MK 111

by

S. Raimondo, A. A. Haasz and B. Etkin

SUbmitted September,

1978

Acknowledgements

The authors grate~u11y acknowledge the contributions o~ Messrs.

C. G. Barringer, J. Fairgrieve and A. Basacchi to the development o~ In~rasizer Mark 111. The authors also wish to thank Dr. R. T. Woodhams ~or providing the incentive ~or the investigation o~ mica. The ~unding o~ the research was

provided by ~rasizer Limited, the McAllister Foundation and the NSERC

under grant No. A9188 and A0339.

'"

"

.

Abstract

This report details the research and development of a horizontal laminar flow particle classifier (Infrasizer Mk lIl) which is capable of si zing particles of a few ~m to hundreds of ~m according to the following aerodynamic principle. Particles of different terminal velocities, when introduced into a steady horizontal laminar air stream, will experience different trajectories, resulting in particle sizing based on terminal velocities (which in turn are functions of particle density, size and

shape). Maj or effort has been expended on: (l) the attainment of the low turbulence uniform velocity flow field, and (2) the development of a feeder apparatus used for introducing the powders into the air stream

with minimumparticle interference effects. Our work has led to the design and construction of various prototypes which have been tested on various materials including glass spheres, mica, iron ore/silica mixtures, and

1. 2.

3.

4.

5.

COm;ENrS Acknow1edgements Abstract Notation IN.rRODUCTION

THEORErICAL CONS]J)ERATIONS FOR HORIZONrAL ELUTRIM'ION 2.1 E~uations of Motion for an Independent Partic1e 2.2 Simp1e Mathematical Model

2.3 Terminal Velocity Calculations 2.4 Second Mathematical Model 2·5 Partic1e Interaction

PROTCY.rYPE DESIGNS OF OO'RASIZER MK III

3.1 Wind Tunnel Design

3.2 Design of Feeder Apparatus EXPERIMENI'AL RESULTS

4.1 General Observations

4.2 Sizing of 325 Mesh Particles

4.3 Spherical G1ass Beads 4.4 Mica 4.5 Density C1assification CONCLUSIONS REFERENCES TABLES FIGURES iv Page ~ ii iii v 1 1 2 3 3 6 7 8 8 11 13 13 14 15

16

18

20 22

Notation

Cd drag coefficient C

fe dumulative fraction weight distribution on tunnel floor

Cwfe(x) normalized cumulative iron are weight distribution on tunnel floor Cwt (x) normalized cumulative weight distribution on tunnel floor

d particle diameter

'-d

a particle projected diameter !d drag force vector

F

-g gravity force vector g gravity constant

H height dropped by particles, usually height of tunnel test section h constant defining an apparent fee ding source to take into account the

acceleration length of the particle

h _o the value of h when the particle initially enters the tunnel wi th terminal velocity

L landing posi tion for the parti cle mp mass of particle

Re Reynolds number

S cross-sectional area of particle U relative velocity of particle to flow Uo mean tunnel velocity

ui initial velocity of particle Vol volume of particle

'

..

V velocity vector of fluid V velocity vector of particle -p

Y

_R relative velocity vector of particle to flUid VR magnitude of relative velocity vector

V_t W fe W_t Wfe(x) Wt(X) (x,y) ex v,a b.d

particle terminal velo city

total weight of iron ore content in sample processed total weight of sample processed

iron. ore weight distribution along tunnel floor

weight distribution of sample processed antunnel floor cartesian coordinates of particle position

volume shape coefficient

difference between largest and smallest predieted diameters

difference between large st and smaLlest experimentally determined diameters

e

angle of flat splash plate Il fluid viscosity

p fluid density

Pp particle density

1. INrRODUCTION

In the ear1y part of this century, the University of Toronto had been invo1ved in the c1assification and sizing of sub-sieve particulate matter. The work of Professor H. E. T. Haultain (Ref. 1) resulted in the deve10pment of the Infrasizer Mk I* which em:p1oyed vertical air elutriation. The technique, though working quite well down to 10 J..Illl, does have a mmlber of 1imitations, one being the excessive experimentation time and another being the restriction to seven fractions.

With these limitations in mind, research in this area using horizontal e1utriation has been reinitiated at the University of Toronto, Institute for Aerospace Studies (UTIAS). In the past horizontal elutriation has been used for dust collectors (Ref. 2), air samplers (Refs.

3, 4),

and for measuring drop si ze distribution in rain (Ref. 5). Our work resulted from tbe natural extension of the wind tunnel simulation of rain penetrating tbrough airjets, which involved the use of miniature spherical glas.s beads to simulate the precipi tation. The nature of the precipitation feeding system required the use of the wind tunnel air velocity to bring the particles over the test model, as wel1 as to minimize the possib1e c1umping of the particleso The size filtration or segregation of the particles which resulted became a parameter of concern in the precipitation work; however, the possibility of using the technique to classify the particles into size groups was demonstrated.

The initial investigations performed by Barringer (Ref. 6) demonstrated that in a laminar flow, with spherical particles in the range 50 to 150 Ilm,

c1assification could be achieved and predicted with some accuracy using a simp1e theory 0 Since then tbree prototype tunnels for the Infrasizer Mk III and four feeders have been designed and tested. Uses for the infrasizer are envisioned for both research and industry. It may be used in research to obtain very close size ranges or "monosized" particles required for f'lllTther experimentation or chemical analysis. As a laboratory tool the infrasizer can be used to obtain siz.e distributions when it is used in conjunction with microscope analysis or calibration tableso In industry uses may range from size c1assification by replacing screening to density classification which can be accamplished in conjunction with screening.

2. THEORETICAL CONSIDERATIONS FOR HORIZONTAL ELurRIAXION

In horizontal elutriation, as it is carried out in the Infrasizer Mk lIl, the classification or si~ing of particles is accomplished by dropping the particles into a horizontal, laminar airstream, contained within a wind tunnel, and cOllecting them on the floor in narrow trays. In general, the final position of the particles on the tunnel floor depends on the time history of aerodynamic, buoyancy, gravitational and inertial forces acting on the particle during i ts flight. Larger, denser material will have larger falling veloci ties and thus land farther upstream tl1an smaller, less-dense material. The advantages of horizontal elutriation over vertical e1utriation can now be seen. In horizo~al elutriation sizing is complete once the partic1e lands on the floor; whereas, in * Tbe Haultain Infrasizer, consisting essential1y of seven vertical cones through which the air flows in series, has been designated the Mk I. The Mk II is an impraved version of the Mk Io The present development, based on an essentially different concept, is termed the Mk lIl. Infrasizers are manufactured and

verticaJ.. elutriation, the limiting size which can theoretically leave the tap of the elutriation chamber has its gravitational force just offset by the aerodynamic force, and it could take many hours for complete classification. The number and position of the classification cut points in the Infrasizer Mk III are completely variable by appropriate choice of collection trays. 2.1 Equations of Motion for an Independent Particle

Simplified equations, in two dimension, are derived based on a particle having mass ~, suspended in a laminar flow and subject only to aero-dynamic drag, gravitat~onaJ.. and inertial forces. Assuming the particle is moving with velocity ~

=

(~, vp), relative to an inertial frame, through a fluid of velocity V = tu,

v),

then the particle velocity relative to the fluid

is -with

v

=

V -

Y.

=

(v

up -

VU)

. -R -p P

v

=

Iv I

=

(u - u)2 + (v _ v)2]1/2 R -R P P

The aerodynamic force acting on the particle is

(2.1)

(2.2)

where Cd is the drag coefficient and S is the particle' s frontal area. The gravitational force, corrected for buoyancy, is

F

=

m

(~

). (2.4)

-g :-g

where m

=

(1 - p/pp)IIlp' Application of Newton' s Second Law yields

or mp (:p )

=

-~

+

~

1

= -

P C S 2 d p

To solve the differential equations (2.5) we require the dreg coefficient, which is a nonlinear function of Reynolds number and hence of particle velocity. It is then found that the equations have to be solved numerically.

2.2 Simple Mathematical Model

As a first analysis, the following simplifying assumptions can be made:

(1) The flow is laJDinar and uniform in the horizontaJ. direction, with no transverse components and no boundary layer growth along the tunnel waJ.ls, that is, u = Uo and v =

o.

(2) The acceleration length for the particle to reach the fluid velocity Uo and its terminal velocity Vt relative to the fluid are negligible. Thus equations

(2.5)

degenerate to the simple relationship

(2.6) where (x,y) is the locus of points mapping the partiele trajectory. Af ter the particle h~s fallen from a height Hit has landed at a distance L along the floor given by

Barringer'in his wórk used spherical glass beads of density 2.42 and

4

g/enl3 in the si ze range 50 to 150 Jlnl.. He found the above relationship to hold quite well for the smaller particles but overestimated the distance L for larger particles. He further attributed the aQomaly entirely to particle inter-action.

2.3

Terminal Velocity Calculations

To apply equation

(2.7)

the terminal velocity of the particle must be evaluated. A particle falling under the force 'of gravity in a motionless fluid will attain a constant terminal velocity Vt when the net (corrected for buoyancy) gravitational force equals the resisting drag force acting on the particle. Thus from equations

(2.5),

setting up =

0

and vp. = -Vt yields

. = [ 2g Vol( P_{p -} p) ] 1/2

V_t SC '

P d '

(2.8)

From the above equation, i t can be seen that the terminal velocity is based on a compli cated relationship of particle parallleters • It is not only dependent on the density and volume to cross-sectional area ratio of the partiele, VOl/S, but it aJ.so depends on the drag coefficient which in turn depends on the shape

and Reynolds number effects, see Lapple (Ref. 2) and Allen (Ref. 7). (a) Spherical Particles

For spheres, geometrical relationships re duce equation (2.8) to

V

=

P

[

4g d( p - p) ] 1/2

t 3C

For a sphere fal1ing in a f1uid where the viscous forces daminate, Stokes theoretical1y demonstrated that the drag is proportional to d·Vt and found

that the drag coefficient is given by Cd = 24/Re. By substituting in Eq.

2.9

(2.10)

The above equation is valid for Re

<

0.1 and is known as Stokes' Law. For very smal1 particles whose si ze :i,s of the same order of magnitude as the mean free path of the molecules of the f1uid, the sett1ing speed wi11 be

somewhat greater than that calculated by Stokes' Law. The sett1ing velocity

as ealculated by Stokes' Law should be multiplied by a factor known as the Cunningham Correction Factor to obtain the actual sett1ing velocity. The

correction is only appreciab1e for particles 1ess than 5 ~ for sett1ing

in gases. Smal1 particles suspended in a f1uid, in addi tion to a directed motion as a result of an ëxternal force, are also subject to a random

motion known as Brownian movement as a result of collisions with individual f1uid molecules. Brownian movement becomes appreciab1e for particles 1ess

than

3

~ in diameter and is predominant for particles smaller than 0.1 ~.

At much 1arger Reyno1ds numbers (1000

<

Re

<

200,000) turbulent flow around the sphere occurs, the viscous forces become neg1igib1e, and

the drag coefficient is sUbstantial1y constant with an average Cd = 0.44 or

[

gd( p - p)

J

1/2

V_t

=

1.74

~

(2.11)

In between the above two regions, 2

<

Re

<

1,000, the drag coeffieient can be approximated by

(2.12)

or

(2.13)

This equation is tedious to evaluate and i t also introduces considerab1e error since the drag coeffieient equation (2.12) is a particular1y poor

approximation . Better approximations do exist for narrow ranges of Reyno1ds

nuIDber, but they are usual1y expressed as a short series invo1ving the summa-tion of terms with Reyno1ds number raised to different powers. Thus to obtain the termináJ.. velocity direct1y from the Cd vs Re re1ationship an iterative approach must be used since the drag coefficient invo1ves a velo city term.

To avoid iteration, it has been Fointed out that the term(CdRe2) for a sphere fal1ing at Vt is not dependent on Vt, and ean be eva1uated by the expression

--- .

.

"'

.

..

2 4gPd3(p - p) (c Re )

=

p Cl: 31-12 (2.14)

From a graph of' (CdRe2) plotted

ag~nst

Re, the Reynolds number can be obtained and the terminal veloei ty deduced. Similarly, if the terminal velocity is known and the diameter required, the term (Cd/Re) which is independent of' size can be evaluated using

4g(p - p) (Cd/Re) =

~

3

3p t

(2.15)

The Reynolds number and thus the diameter can be obtained f'rom a graph of' Re plotted against (Cd/Re). Heywood (Ref. 8) suggested a similar technique but instead of' plotting the graph he used log-log tables f'or the respective terms. Figure 1 shows the terminal veloei ties of' spherical partieles

f'alling in air, and Table 1 gives the Cd, (Cd/Re) , (CdRe2) and Re relation-ships.

(b) Irregular Shaped Partieles

The data on the settling veloeities of' irregular shaped partieles are not very complete and are further camplicated by the variety of' methods f'or expre~sing the partiele shape and size. Two of' the mos~ promising techniques for reducing the data were proposed byHeywood (Ref'. 8) and

Wadell (Ref.

9).

Wadell proposes the redUt;:tion of'the Cd vs Re relationship in terms of the partiele's sphericity, which is defined as the ratio of' the surf'ace area of' the sphere with the equivalent volume to the total surf'ace area of the partiele. A reproduction of his results can be seen in Fig. 2. The dif'ficulty in using the sphericity as an appropriate shape factor arises f'rom the difficulty of' evaluating the surface area of' the partiele which is not easily available f'rom standard measuring techniques. Heywood proposes the use of the volume shape coeff'icient which relates the volume of' the irregular partiele to the volume of' a sphere having the same projected area as the irregular partiele , i. e. ,

Vol = ex d 3 v,a a

Heywood's results, which he presented in the f'orm of' corrections to be added to the terminal veloeities of' spherical partieles, have also been presented by Allen (Ref'. 7) in a more convenient form, Table 1, by providing values of' log (Re2Cd) and log (Cd/Re) f'or four values of'

<:Xv

a

=

0.1, 0.2, 0.3, 0.4. Allen f'urther reduced the data to Cd - Re relatibnships f'or the four volume shape coefficients. This data is sui table for numerical evalua-tion of' Eqs. 2.5.

Further camplications arise if' the partiele is within Stokes'

Law regime where the partiele will in general retain its initial orientation, and also, asymmetrie partieles such as ellipsoids and discs do not fall

vertically but tend to drif't to one side unless they are dropped wi th a princi:pal. axis of' symmetry parallel to the gravity f'ield. Unsymmetrical bodies may not attain steady motion but rather may exhibit spiralling or wobbling.

The special case of a disc falling in Stokes' Law region in two orientations is useful to examine. The fOllowing relationships are

applicable for a disc falling with its face horizontally,

and

For a disc falling edga on

and Re

=

dUp/1l 16 F = - IlUd d

3

(2.16) (2.17)

Fram Eqs. 2.16 and 2.17 one can conclude that the terminal velocity of a disc is proportional to the product (d·t) and the ratio of the terminal velocities of a disc falling in its two extreme orientation is only 1.5 to 1.

2.4 Second Mathematical MOdel

Experimental observations of particle trajectories revealed that the acceleration lengths were ,not entirely negligible. A new model for the partiele trajectories is proposed to take into account the acceleration length and predict the correct landing point. It is observed that af ter the partiele has accelerated to the fluid velocity and is falling, the trajëctory has asymptoted to a trajectory parallel to the one predicted by Eq. 2.6. One can define an apparent new source for the particle located h below the actual source by extrapolating back along the line coincident wi th the: asymptote, Fig. 3. Thus, the landing point can be predi cted by

\

and is valid for H!

_.

_' ~ l5h •

U

L

=

~ (H - h)

V

t

(2.18)

To obtain values for h, Eqs. 2.5 were solved (Ref. 10) for the case of spherical particles (pp

=

2.4 g/~3) released into an airstream

,

.

of constant horizontal velocity with an initial downward velocity equal to i ts terminal veloei ty. The traj ectories of the partieles can be seen in Figs. 4 and 5. The coordinates .of the point at which the partiele reached the airstream velocity was noted and h was plotted as a function of Vt2 yielding the straight lines shown in Fig.

6.

The source height corrections h for a 100 ~ diameter partiele were also evaluated for various air stream veloei ties, see Figs. 7 and 8. Another point of interest was the investigation of the initial partiele velocity (Up= 0, Vp= ui) on the acceleration length. The results are shown in nonQl.lD.ensional form, i.e., (h/ho) vs (Ui/Vt), where ho is the value previously obtained for ui = Vt, see Figs. 9 and .10. By observing the trajectories plotted in Fig. 11 it can be seen that a spread in the landing positions can arise for a spread in initial veloeities. The trajeetory which most closely approaches the one predicted by the simple theory occurs for aparticle entering the airstream at rest (Ui = 0). With respect to the Infrasizer Mk 111 apparatus, there are two ways in which varying ini tial conditions can arise in the experiments: (1) improper feeder design and (2) the effect of partiele concentration.

2.5 Partiele Interaction

A large discrepancy in the 'quality' of slzlng obtained in the '

e~eriments can be caused by partiele interaction. The first and most

obvious way is if two or more partieles are held together through surface forces which may be frictional or electrostatic. The agglomeration of

~articles will now act aerodynamically as a single larger partiele and fall faster. The second type of interaction is purely aerodynamic in nature. It has been observed that a small partiele in the wake of a larger partiele will fall faster than its calculated terminal velocity. It can,in fact, travel with such a velocity that it is able to overtake the larger partiele, or else, the larger partiele may piek up the smaller partiele so that it revolves as a satellite. The effect of this type of interaction will manifest itself in the Infrasizer results by finding tiny partieles where only very large ones are expected. Similarly, when a group of partieles are in close proximity to each other to form a cloud or a cluster, the terminal velocity of the cluster is higher than the terminal velocity of the individual partieles. The resulting effect on the Infrasizer is that the partieles may settle without any classification taking place.

Quantitative analysis to determine the effect of interpartiele distanees andjor concentration of partieles on the interaction is scarce and limited to Stokes' Law Region. Rappel and Brenner (Ref. 11) discuss the interaction of several partieles and show that two partieles falling as close as 10 diameters-_apart exhib;i t negligible interaction • Boardman and Kaye (Ref. 12), who have measured the largest veloei ty atpentation, reported that at between 0.2 to

3%

concentration by volume, 900 ~ diameter spheres falling in liquid paraffin have veloeities as much as one and a half times greater than their Stokes velocity. At 0.1% concentration, which corresponds to an average inter-particle spacing of 8 diameters, velocity augmentation was already showing up. The interaction which may result at higher Reynolds number (Ref. 13) can only be speculated at, but it is believed that it becomes more pronounced as Reynolds number increases. One can only state that any type of interaction will have a detriment al effect on the quality of sizing that is achieved and should therefore be minimized.

3 . PRCJ.racYFE DESIGNS OF INFRASlZER MK 111

The research program has been centred on two basic design

aspects of the Infrasizer Mk 111:

(1)

the attainment of a low-turbulence uniform flow, and (2) the development of an apparatus which is capable of introducing the sample into the airstre~ with a minimum of particle-particle interaction. The work has led to the design of three working prototype tunnels and also four prototype particle feeders • .

3.1 Wind Tunnel Design

(a) ~IAS Precipitation Wind Tunnel

The initial experiments performed by Barringer (Ref. 6) were conducted in the pre-existing UTIAS Precipitation Wind Tunnel, Fig. 12, which was designed by Harting (Ref.

14)

for another purpose, viz, stu~es

of particle-airjet interactions. This is a closed circuit wind tunnel with a. square cross-section of about 1 m x 1 m and a test-section length of 5 lIl. It is capable of test-section speeds ranging from

15

cm/ s to

300

cm/ s. The flow is uniform with a few cm thick boundary layer at the end of the test section, and turbulence intensity levels of the order of

0.5%.

The particles to be sized were fed into the tunnel through ports in the roof and were

collected on the floor in a series of trays

2.5

and

5

cm in width.

(b) ~frasifer Mk 111 - Model

001

The first model of the Infrasizer Mk 111, designed by Fairgrïeve (Ref.

15)

utilized a diffuser, in the vertical plane, with a total contained angle of

10°,

see Fig.

13.

The initial cross-section was

'

21.6

cm high and

49.5

cm wide, and the test section length was

150

cm. The concept of using

a diffuser configuration was based on the fact that there exists a nonlinear relationship between the terminal veloei ty, and the 1anding position of a particle. Thus, the decreasing velocity as a function of distance along the test section, resulting from the increasing tunnel cross-sectional area would cause a particle to travel downstre~ a short er distance than. in an equivalent constant cross-section wind tunnel, with the s~e initial velocity and cross-section. As a result, a larger tunnel velocity can be used at the initial cross-section which ~nables the mass distribution for the larger particles to be stretched along the tunnel floor and yet still allow the smaller particles to land within the tunnel. The tunnel was also equipped with a false floor and ceiling such that other diffuser angles as well as parallel sided tunnels colild be investigated. The air-flow was provided by a centrifugal blower W:;:'.ose volume air-flow colild be

controlled via a power trrolsformer controlling a universal electric motor. The inlet of the tunnel had a wide angle diffuser fitted with several 18 mesh screens to avoid flow separation • A package of ten 18 mesh screens was placed between the inlet diffuser and the test section to reduce turblilence\ and provide uniform flow at the entrance to the test section.

The ~articles colild be introduced into the tunnel through a slot cut

transversely into the roof at the beginning of the test section. The 'sized' particles we re collected in thirty trays of

5

cm width and 1.8 cm depth, placed on the tunnel floor. Observation and accessibility of the trays was provided by a hinged plexiglass door which also formed the front side of the tunnel. The tunnel could be used in an open circuit configuration

..

or in the closed circuit mode by connecting the outlet of the fan to the

diffuser inlet via a.10 cm diameter duct.

With the open circuit configuration astrong inward flow resulted at the slot which could not only affect the flow characteristics within the

tunnel but would also ~ive the particles an initial impulse downward,

which would in turn change the landing position by reducing the effective tunnel height. In a closed circuit configuration no flow exists through the slot if the slot is the only opening in the tunnel. However, any tunnel leaks upstream of the slot will result in an inward flow with the conse-quences described above, whil@ leaks downstream of the slot will cause an outward flow which could prevent the smaller partieles from entering the tunnel. Thus, the tunnel nrust be well sealed to eliminate all leaks.

Because of the disadvantages associated with the open circuit configuration, further experimentation was only performed with the closed circuit arrangement. Velocity profiles and turbulence intensity levels were obtained using a hot wire anemometer with temperature campensation

for a number of points at the slot cross-section and at a position 1.1 m

downstream of the slot for both the parallel walled tunnel and the 10° diffuser. At the initial cross-section the uniformity was reasonable with

a 5% standard deviation, but thè turbulence intensity level was about

3.CJ'/o.

The above values were for the centre portion of the tunnel and higher values were recorded in the boundary layer where both non-uniformity and turbulence

intensity levels exceeded 10%. It is also interesting to note th at the

volume flow deficiency in the boundary layer did not result in a slight increase in the mean velocity across the tunnel. Rather, it resulted in "jetting" of the flow just outside the boundary layer. At the cross-section 110 cm downstream of the slot, for the parallel sided tunnel the turbulence

intensity increased to about 6% at a position

7

cm above the collection

trays. The increase in the turbulence levels were the re sult of trays

tripping the flow. In the 10° diffuser configuration at the same

cross-section the velocity profile was not uniform as would be expected, however,

the turbulence intensity level now reached 35% at points

7

cm above the

trays. Since the turbulence level in the diffuser configuration was

excessive, the tunnel was used onlywith parallel walls for the subsequent experimentation.

(c) Infrasizer Mk III - Model 002

With the concept of a closed circuit and parallel walled tunnel in mind, Model 002 was designed with the aim of achieving a greater flow uniformi ty, a higher mean velocity (100 cm/ s) and a turbulence intensi ty

level under 1%. In Model 001 when the tunnel velocity was at maximum

: (38 cm/s) the pressure drop across the fan was 90 mm of water and more

than 95% of this pressure loss occurred in the small diameter return duet. The overall dimensions of Model 002, Fig. 14, are: width 45.5 cm, height

40.5 cm, and length 220 cm. These dimensions encampass a test section of

24 cm height, 39.4 cm width and 122 cm length, and a return duct of 10.2 cm heigllt and 39.4 cm width. The estimated pressure loss of the circuit was 6 mm of water. The fan selected for the tunnel was a Tarzan TN3C2 (Rotran Manufacturing Company), an electronic camponents cooling fan, 17.5 cm in diameter with a shaded pole electric motor allowing for variable speed via a solid state speed controller. The specifications for the fan are

a capacity of

10

cubic metres per ndnutefor free flow and

23

mIn of water

stagnation pressure. To obtain uniform flow a screen package consisting of tEm screens was placed 15 cm upstream of the feeding slot and consisted of (in the direction of the airflow): five

18

mesh, four

30

mesh and one

40

mesh screens. To help turn the flow in the corner immediately preceding the screen package three

18

mesh screens were used as.· shown in Fig.

14.

The tunnel layout was designed by Basacchi (Ref.

16)

,·

and he later performed experiments to obtain velocity profiles for several screen combinations, ·

with a honeycomb inserted and with rounded corners. Vertical profiles were obtained at the slot cross-section and at cross-sections

30, 61

and

122

cm

downstrerun from the slot. .

The effect of the various screen combinations on the flow ware compared in terms of vertical velocity profiles obtained midway between the two sides of the tunnel at the slot cross-section. The velo city profiles were obtained at maximum volume flow and the mean velocity ranged from

140

to

180

cm/s depending on the pressure drop provided by the screens, Fig.

15.

In camparing the effect of the screens a turbulence intensity will be quoted and it is to be understood that this value was for the most part uniform in the central region, that is outside the boundary layer. With the screen package consisting of only two screens, a

20

and a

40

mesh, the pressure drop provided by the screens was not sufficient and the flow was 2Cf1/o faster near the roof than near the floor, and the turbulence level was about ~%.- By inserting a

30

mesh screen between the

20

mesh and the

40

me sh· the flow deficiency near the wall disappeared but now a slight jetting near the tunnel floor resulted, and the turbulence intensity was reduced to 1.4%~ Replacing the

20

mesh by another

30

mesh screen the jetting was reduced slightly and the turbulence intensity was reduced to

1.2%.

The designed screen package, consisting of

10

screens, produced the most uniform profile wi th the large st deviation from the mean of only

4%

except for the

j etting near the floor which was as large as lCf1/o; the turbulence intensi ty f or thi s case was

0.6%.

By helping the flow turn the corner, viz, by

replacing the turning screen with a turning plate, the turbulence tntensity

was reduced to

0.5%.

.

Velocity profiles were also obtained to detect the boundary layer growth along the tunnel. Vertical profiles were obtained for a mean tunnel velocity of

140

cm/s at eross-sections

30, 61

and

122

cm downstrerun of the

slot cross-section. The co1lection trays were turned over to provide a flat floor and avoid tripping of the flow, see results in Fig.

16.

It

should be noted that the profile obtained at the

122

cm cross-section is only

15

cm upstream of the fan and some fan influence might be present. The turbulence intensity level along the centreline increased from

0.5%

at the .slot to

0.95%

at

30

cm, it then decayed to

0.6%

at

61

cm and increasdii again to

0.7%

at

122

cm. The turbulenee intensity level in the boundary layer is more revealing. At a height

2.5

cm from the floor the intensi ty increases from

0.56%

to

1.1%

to

2%

to

2.9%

as one proceeds downstrerun from the slot for the cross-sections mentioned above. These tunnel velocity characteristics are within the design specification; however, it must be remembered that the trays are inverted. Once the trays are set in place and the 'flow is tripped by the edges then the boundary layer growth and .

turbulence intensity can both be expected to increase as one proceedsdown-strerun.

(d) Infrasizer Mk 111 ~ l~~el 003

Model 003, Fig. 17, was designed to obtain even better flow char-acteristics. To improve the flow, the return duet was made the same size as the test section (29.8 cm wide and 19 cm high). A 15 cm space was

provided between the test section and.the return duet to allow for a larger

turning radius at the ends. Provisions for a second fan were made in order to double the available pressure drop and allow for a filter to be installed at the end of the test section in order to prevent recirculation of the very

fine partieles • Initial results revealed that .the turbulence intensi ty

level has been reduced at the

slot

,

cross-section to 0.2%. The effects of

the trays (not inverted) on the flow show a boundary layer growth of a few cm by the end of the test section wi th a turbulence intensi ty in the boundary

layer (at a height of

5

cm above the trays) of up to 20%. Further

experimen-~ation with this tunnel is in progress.

3.2

Design of Feeder Apparatus

(a) Feeder - 1

The feeding system used by Barringer (Ref.

6),

as shown in Fig.

18, consisted of a circular cylindrical hopper which contained a sample of spherical glass partieles. The glass beads emerged, under the force of

gravity, through the hopper's base cone via a small orifice, and the partiele mass flow rate was controlled by interchanging base cones with different

size orifices. The glass bead beam flowing from the hopper was allowed to strike one or two flat "splash" plates, which could be adjusted to a wide range of positions and angles in order to produce a "Cluasi-line source". The entire system was enclosed in a plexiglass case so that the partieles would not be disturbed by air currents.

This feeder functioned Cluite well with spherical glass beads but when mineral samples and pcwders of non-spherical shape were placed in the

hopper the particles would not emerge. In order to drive the particles

through the orifice, aer~tion was introduced by (1) a perforated tube

down the centre of the hopper, and (2) through the inside walls of the cones; however, neither techniClue worked sufficiently well on all samples.

The major drawback was that a few particles of approximately the same si~e

as the orifice would get lodged in the base cone, thereby plugging it. This problem could be alleviated by pre-sieving through a coarse sieve

(70 mesh) .

The very tiny powder-like particles also presented a problem, since upon being poured into the hopper the particles near the orifice were compacted by the weight of the rest of the sample, which might have resulted in a bridge structure which would not allow the particles to flow. To ensure that it was not electrostatic forces which prevented the particles fram emerging, the aerating fluid was humidified to 50% relative humidity;

however, no noticeable effect was observed on the flow. It was also found

that vertical alignment of the hopper was critical if any particle flow

was to be achieved. Aeration alone did not seem to be sufficient. Vibrating the hopper, when the sample had been presieved to remove the larger partieles, did allow the particles to emerge through the orifice, but now, when the

stiGking could be partial1y removed by also vibrating the plate and setting the plate at a. very steep angle, aJ.most vertical. This, however, did ,not allow for asufficiently wide line source and consequently not sufficiently low particle concentration to eliminate inter-particle interaction. The overall concept of the cylindrical hopper with the beam of particles impinging on a flat 'plate was ab andoned , not because i t could not be made to function as desired, but because the amount of time required to investi-gate all the possible parameters and the complexity of the device were excessive.

(b) Feeder -

g

Af ter investigating anumber of aerodynamic turbulence air chambers without success, a new and totally different feeder concept was arrived at. The hopper, see Fig., 19, consisted of a triangular trough that was lined with three replaceable sieves, of either 50 or 70 mesh. The trough was 30 ~ in length and the lower sieve defined a line source

1.5

mm wide. The hopper was allowed to slide on two mounting rods and was tibrated via a reciprocating arm mounted ecëentrically wi th a 0.8 mm offset on the shaft of an induction electric motor. The vibration was in the horizontal plane and perpendicular to the airflow direction.

Feeding the hopper was attempted in two different ways. The first was to pour the entire sample, usually 50 or 100 gram, into the upper screen and then start the hopper vibrating. It was observed that the particles quickly passed through the upper screen, and it is believed, just as quickly through the middle screen, although direct observation of this was not possible. The particles would th us accumulate above the lower screen. The lower screen became overloaded which resulted in a very slow mass flow rate, or in the worst cases prevented the particles from ever penetrating the screen., The second method of feeding the hopper was to start the hopper vibrating and then sprinkle the sample into the

hopper. The time taken to run the experiment in such a manner (cleanup and analysis time not included) was approximately 20 seconds to just under five minutes depending on the material used. '

(c) Feeder -

3

Feeder -

3,

see Fig. 20, though similar to Feeder - 2, used a different method of supporting and vi brating the hopper. The hopper was supported on two rubber studs, one' at each end, which allowed for lateral vibration. An electric motor was also mounted between the rubber

supportsand the ensuing vibration was the result of an eccentric weight on the 'shaft of the motor. Further changes included the provision of only two removable screens in the hopper, and the line source was increased to

6

mm in width. To better define a line source one of the lower edges of the hopper was later extended to provide a "splash" 1>Late for the partieles before entering the airflow. To eliminate the need for the operator to sprinkle the sample and to make the feeding rate more uniform, a split tube with a. length equal to the hopper length was designed to contain the sample above the hopper. The particles were introduced into the hopper by rotating the split tube (which was mounted on the vibrating supports) slowly via syncbronous a-c motors (in the speed range 1/30 to 2 rpm). These motors could quickly be interchanged to obtain a desired feed rate.

Initially the split tube consisted of a circular tube cut lengthwise and the material spilled over the edge. It was noted that this did not provide for a uniform feed rate, and the cross-section was modified in such a way that the material would spill fram the centre of the tube and yield a uniform feed rate.

(d) Feeder -

4

Feeder -

4,

although similar to the previous two feeders, was more compact and was provided with a split tube giving a uniform feed rate, see Fig. 21. The particles leaving the split tube af ter penetrating a single screen were purposely made to strike both sides of the trough which formed the splash plates. With such a compact design the feeder vibration motor could no longer be mounted on the vibrating supports and thus had to be mounted on the base plate, and the vibration was once again provided using a reciprocating arm. The split tube was rotated via a non-slip pulley which was driven by an electric motor througha small multi-speed transmission gear box allowing for the feed rates to vary.

4.

EXPERIMENTAL RESULTS 4.1 General Observations

In this section general observations on the quality of sizing will be made. The experiments discussed are for the most part qualitative in nature with the contents of a few trays being observed under a micro-scope.

The initial results presented were obtained using Feeder - 2 and tunnel Model 001 combination; for tray sizes and location of trays on the tunnel floor, refer to tray pattern A in Fig. 22. The glass beads of Fig. 23 had a true specific gravity of 2.42 g/cmj _{and were in the 80 to 150 ~m}

size range. The beads were poured into the hopper with the 70 mesh screens in place, the tunnel was turned on (~20 cm/s), and the vibration of the hopper started. The duration of the experiment was less than 1/2 minute for a 50 gram sample. As can be seen from the photographs, the sizing is quite good and the glass beads were later used as tracers to indicate the quality of sizing in non-spherical materials.

When the experiments were performed using cower concentrate as the working sample it was found that the entrainment of small particles by larger oneswas significant, Fig. 24. This problem was alleviated by pre-sieving the sample through a 270 mesh sieve (53 ~). The experiment using copper concentrate was then repeated using the +270 mesh particles with a tunnel velocity of

36

cm/s, and it took a few minutes to feed through an 18 gram sample, see Fig. 25 for results. For particles greater than 53 ~ it was found that the accuracy of si zing could be improved if the contents of a tray was recycled several times, Fig.

26.

It was also found that when sub 53 ~m particles were recycled those which had initially landed in trays No. 16 through No. 20 now landed in the first 10 trays. This

indicates that . clustering may be a problem and should be avoided. Similar qualitative results obtained in the Infrasizer tunnel Model 002 (tray pattern B, Fig. 22) with Feeder - 3, and a tunnel velocity of 135 cm/s can be seen in Figs. 27, 28 and 29.

· 4.2 Sizing of 325 Mesh Particles

One of the aims of our research is to extend the lower limit of the sizing down to about 10~. In dealing with very small particles. it must be realized that the effects of clumping and aerodynamic interference between the particles are amplified. Experiments were performed for two types of materials with sizes less than 44~: glass beads and mica. The experiments were in both cases conducted in tunnel Model 002 with Feeder - 3 (tray pattern B, Fig. 22). The feeder apparatus included a splash plate which was attached to the hopper,. 140 mesh (105 ~m) screens used in the hopper trays, and the split tube (with a non-uniform feeding rate) was rotated at 1/30 rpm.

During the experiments with the glass beads the two conditions which had beenpostulated regar.ding the interference among the particles were observed: (1) the dependence on the numoer density flux of particles, (2) the dependence of the critical number flux on the air speed. At the lowest achievable tunnel speed, 12 cm/s, the interference between particles was observed as "turbulence" or "waviness" as the glass beads fell to the tunnel floor. This is attributed to clustering of the spheres or' an increase in the number density flux. It was only at very small mass flow rates that the apparent interference disappeared. When the experiment was repeated at higher tunnel velocity and approximately 1 gram of beads was placed in the feeder, the interference could initially be seen but as the mass rate decreased, as the feeder emptied, the interference also decreased. A correlation could be inferred between the quality of si~ing and the observed interference by observing the beads under a microscope. In another experiment the tunnel speed was set to.44 cm/s and a very small feed rate ~ 0.02 gis was used to obtain the results of Fig. 30. The quality of the sizing is not as good as that achieved for the larger particles, see Section 4.3.

The mica 325H material visually appeared to be co~osed of tiny powder-like particles. The initial experiments were, therefore,conducted using the slowest attainable mass rate (1/30 rpm feed rate) and the tunnel at the slowest possible velocity, 12 cm/s. A screen was also placed in the hopper, anticipating some clumping. The clumping did result and the particles balled up into spheres of about 1 IlDll in diameter, and when some material did penetrate through the screen it was classified according to the size of the elump.

Severaltechniques were invéstigated to remove the static charge, believed to be the cause of the clumping. The first was the application of a cOIlDllercially available product, "Static Guard" (in spray form) which proved especiallydifficult to apply and its true effectiveness could not be evaluated. Another method involved the increasirig of the air relative humidi ty content from

<

10% to .~

5Cf'/o,

for both the air circulating in the tunnel and for the samples. The particles now proceeded through the screen more readily and when they entered the tunnel (set at the slowest speed) they travelled downstream .. within a vertical distance of two inches from the tunnel roof which seems to indicate that the particles were now falling independently of each other. Since the lowest tunnel velo city was not sufficiently low to cause the particles to land on the floor, the flow in the return duct was choked by a flat plate with a slit in it. The flow

14

speed was thus reduced to the same order of magnitude as the terminal veloci ty of the particle. The turbulence levels in the flow caused by the choking plate were such that the trajectories of the mica platelets were not straight lines, but rather, reseIDbled smoke diffusion.

To use the Infrasizer Mk III to classify these materials of very small sizes requires more research into the feeder designs. 4.3 Spherical Glass Beads

Q,uantitative analysis was perlormed for two samples of glass beads with Pp = 2.42 g/cm

3 •

Sample I consisted of particles in the size range 120 to 350 fJlD. and Sample II was in the range 50 to 175 fJlD..The samples were sized in the precipitation wind tunnel using Feeder - 1. Samples I and II were run with tunnel speeds of 178 cm/s and 60 cm/s, and at mass feeding rates of 0.13 gis and 0.17

gis,

respectively. The beads were collected in trays arranged on the floor as indicated by tray pattern C in Fig. 22. The contents of the trays were observed and photographed using a microscope ; see Figs. 31 and 32. The average particle diameters (Column 4) and standard deviations (Column 5), as well as 'the terminal velo city (Column 6) based on the average experiment al diameter, and the predicted average diameter (Column 7) based on the simple model are given in Tables 2 and 3. The predicted diameters overestimate the sizes of the particles found in a given sample and this is a direct consequence of

neglecting the acceleration length for the particle. As a first correction the values in Column

8

were calculated using Eq. 2.18 (Second Mathematical Model) in conjunction with Fig.

6

which assumed that each particle was introduced at its terminaJ. velocity • Discrepancies between the diameters thus predicted and the experimentally observed values can still be found and for Sample I the predicted values over-correct for the larger particles while the correction is too small for the smaller ones. A possible explana-tion for this anomaly may be based on the fact that the beads had fallen a distance of about 10 cm under the force of gravity (assuming negligible aerodynamic drag) as a beam before striking the splash plate. At impinge-ment on the plate all the particles would be travelling with a velocity

v'

=

.Jgs/2,

s being the di stance of travel ; for s

=

10 cm, v' ~ 70 cm/ s •

If it is further assumed that each particle strikes the plate independently of each other and retains all of its momentum af ter scattering, then the vertical velocity component ui with which all the particles enter the wind tunnel is v'cose where e is the angle of inclination of the lower splash plate (e

=

20° was used for the calculations). The landing positions for the particles in Sample I were calculated using Eq. 2.18 in conjunction with Figs. 10 and 6 and are shown in Column 9 of Table 2. Similarly, the landing points for the particles in Sample II were calculated from Figs.

9

and 6 and are shown in Table 3. These predictions are now better for

Sample I but are only marginally better for Sample II. Thi s analysi s shows the discrepancies in predicting the landing positions of the particle and can be explained in terms of the acceleration lengths and initial condition for the particles (in a previous study, Barringer attributed such anomalies to terminal velocity augmentation, Ref.

6).

Two values to indicate the quali ty of the separation are also shown in Column 10 and Column 11 of Tables 2 and

3.

Since the trays are of fini te width a range of sizes can be expected to land in a tray and

the difference ~d between the large st and the smallest predicted diameters, using Eq. 2.7, can be defined. Column 10 is the ratio of Me:M where Me is the difference between the largest and smallest particles measured in the tray, while Column 11 is the ratio of 2cr/M.

Another quantitative analysis was performed on two other samples of glass beads using tunnel Model 002 and Feeder - 3 wi th only the upper

. screen (100 mesh) in place. The tunnel velocity was set to 120 cm/s and 100cm/s for Samples III and

IV,

respectively. The sized samples,collected in 5 cm wide trays, yielded the results of Tables 4 and 5 and Figs. 33 and

34.

The cumulative mass distributions, Fig.,

35,

show a deficiency in the amount of the larger particles present, which is the result of having the upper diameters of the initial Samples III and

IV

limited by sieves af

140 and 200 mesh, respectively. 4.4 Mica

A considerable amount of time has been spent on the classification of mica and the work has proceeded on two fronts: (1) upgrading of the quality of mica by separating out the gangue or granular material, and (2) obtaining mica samples of varying aspect ratios.

(a) Upgrading the Mica Sam,ple

6o-z

The gangue material in the mica samples consists of granular particles which resulted during the crushing operation to liberate the mica fram the rock. Since sieving was the only operation performed by'the supplier to elassify the initial sam,ple, there was no mechanism by which granules of the same size distribution as the flakes could be removed. The Infrasizer

can effectively separate the granules from the flakes sin ce they differ greatly in terminal veloci t i es.

In order to analyze a

6o-z

Suzorite mica sample, the Infrasizer Mk III Model 002 was used in conjunction with Feeder - 2 with all the hopper

screens removed. The tunnel velocity was set at 150 cm/ s and the tunnel floor'lined with 5 cm wide trays, except for the last two which were 15 cm in width; refer to Fig. 22 for tray pattern A. A 250 gram sample was placed in the split tube and rotated at 2 rpm wi th the time taken to feed the sample being 15 seconds. The contents of the trays were weighed and recorded in Table 6. The contents of trays 5 through 14 were then sieved in order to determine the fraction of mica and gangue in each of the sized groups.

Observations under a microscope revealed the differences in shapes and colours between the mica flakes and the granular gangue. This made it easy to identify the individual particles, and thus to obtain an estimate of the percenta8e of mica.flakes in the sample. A quick scan over the results, Table 7, reveaJ.s that in any particular Infrasizer sample the mica is by far thr larger constituent. This observation appears obvious since for the same terminal velocity one would expect the diameter of a granule to be much less than that of a .flake. Further reduction af the data of Table

7

produced Table

8.

This was done in order to produce the fewest possible number of cuts that would be required for industrial processes. Also, for each group only two

sieves were chosen. Table 8 shows that with proper selection of sieves a large percentage of the sam,ple can be recovered above the upper screen

both screens contains no mica at all. The material which is trapped between the sieves consists of both mica and gangue; however, it constitutes only a small fraction of the sample.

Same conclusions on the contents of the material and the effective-ness of upgrading the mica can now be made. The 60-z material as recei yed is estimated to consist of 50% mica, in all sizes less than 60 mesh; further-more, the gangue is of the same sizes. If the Infrasizer alone is used to classif'y the material the cut point should be af ter Tray'No.:'B"

Ui

:':

"

Table 6. This would imply that 57'/0 of the material processed (or 85% of the mica) could be reclaimed as 75% purity mica flakes. It must also be noted that the large granules have been removed and the remaining gangue is very small in size (as compared wi th the mica) for visual detection. In order to determine the quality of the mica sample so produced, comparative experiments were performed on a relatively high grade mica (200-H mica) . It was found th at the 200-H mica also contained about 25%by weigbt gangue, and again screening in conjunction with infrasizing resulted in 40% of the material processed (or 80% of the mica) being recovered as virtually 100% mica, flakes.

(b) Obta;ining Varying Aspect Ratio Mica

The work to obtain varying aspect ratio mica, to be used as a filler for,~einforcing plastics, was conducted by Rod Paterson (Ref. 17) as part of his Bachelor's thesis in Chemical Engineering. Using the Infrasizer Mk III Model 001 in conjunction with Feeder - 2, he separated 60-H grade Suzorite Mica at a very slow rate of 0.5 grams per minute. The mica samples were obtained using two passes in the Infrasizer. For the first pass the powerstat, controlling the air velocity, was set at 50 and the contents of trays w to z were retained, see Fig. 22, tray pattern D. Recycling the contents of the first few trays at a power stat setting of 53 yielded samples q through v. All samples were then analysed to determine the average diameters and average thicknesses of the samples; see Table

9.

The diameters wereobtained fram micro-photographs with an X59 magnification factor, Fig. 36. The thicknesses were obtained by floating a monolayer of mica flakes and then weighing the samples. The thickness was calculated from a knowledge of the mica density and froman assumption of the packing factor for the flakes. Further analysis with the use of sieves yielded the results shown in Fig., 37 and Table 10 for samples q and i'i ~"; Paterson' s

analysis of his results led him to conclude that no apparent patterns in classification had resulted. ' The'following is a re-interpretation of Paterson's results.

There were two factors wh~ch made the results difficult to inter-pret • The first was simply an underestimation of the average diameter in Sample q. To correct for this error a newdiameter was calculated using a weighted average of values in Table 10. To ensure. that this was not an arbi trary change, weighted averages were obtained for the diameters and thicknè,sses :in Samples q and'

r.

Tpe second point which helps in the data reduction ,is to note that the'differenc'e in tunnel velocities between the powerstat setting 50' and 53 was only 10%; therefore, th~re is ,no just:Lfica-tion in concatenating' the Samples w through z to the Samples q to v. A simple analysis based on' Eq.

2.6'

'shows that, Samples w through z should overlay Samples u and 'v. A particle which would land at ~$!tion Xl when the tunnel is set at velocity'_{Uo would land at X2 when the tunnel velocity}

or

The above relationship allows for the calcUlation of an equi-valent distanee in order to compare the resUlts; see Fig.

38.

From this figure it can be seen that both diameters and thicknesses decrease with increased floor position. It also becomes apparent that there cannot be a drastic change in aspect ratio since the larger flakes are also the thicker ones. In fact, since classification is according to terminal velocity, Eqs. 2.16 and 2.17 imply that classification should depend on the product of flake diameter and thickness; this is confirmed by Fig.

39.

Thus the Infrasizer is not expected to classify according to aspect ratio. In theory, since classification is performed according to the product (d. t), one woUld expect that by sieving a single sample one coUld recover flakes of varying aspect ratio. The resUlts of sieving, Table 10, do show some aspect ratio sizing, but also one finds that the average diameter and average thickness of the sieved sample do not yield a :constant (d.t). This anomaly can possibly be explained by a varying Cd based on the

orien-tation for the falling flakes.

4.5

Density Classification

The need to classify according to density often occurs in the m1n1ng industry, in particUlar for the separation of mineral ores fram

silicates. As discussed ahove, mica could be separated from the granules because of the differences in terminal velocity attributed mainly to shape differences. Similarly, two partieles of the same shape and size but of different densities have different terminal veloeities • Classification based on density variation was performed for glass beads of different densities and for a mixture of iron ore and silica; see below.

(a) Glass Beads

A mixture consisting of equal volumes of 2.4 and 4 g/cm3 density solid glass spheres in the si ze range

74

to 105 ~ was prepared, and was run through the Infrasizer tunnel Model 001 with Feeder - 2, at a tunnel velocity of 30 cm/s. The analysis was performed by observing the samples collected in 5 cm wide trays (tray pattern A, Fig. 22) under a reflecting light microscope, Fig. 40. The different density glass beads also had a different index of reflection, and thus the density

4

glass beads appeared very light with a dark ring around the circumference, whereas the density 2.4 beads were darker in appearance wi th a light ring around the beads. The analysis showed that

80%

of the density

4

spheres were recovered in trays 4 and

5

(see tray pattern A in Fig. 22), and

90%

of the 2.4 density material was collected in trays 7 through 10. The photographs show the purity of the samples, with only a few of the other density particles

being present. The overlap between the results was confined to the 6th tray and consisted of only about 15% of the initial mixture, by volume.

(b) Separation of Iron Ore fram Silica

Experiments were performed to upgrade the quality of an iron ore mixture consisting of hematite, magnetite, and silica VGith

spe

ei!:!. .:"_'->:

densities of 5.6, 5.2, and 2.7, respectively. Visual observation revealed that the Infrasizer Mk III did classify the material since the iron ore, black and shiny, could be readily distinguished from the silica, brown in eolour with a rough non-reflecting surface • Quanti tati ve analysi s of the infrasized samples was performed by a neutron activation technique with the use of the University of Toronto Slow Poke Nuelear Reactor.

Two experiments were performed. In the first experiment the ndxture was sieved through -120 + 200 mesh, and the sample was run through the Infrasizer tunnel Model 001 using Feeder - 2 at a tunnel speed of 30 cm/s, and 5 cm wide collection trays (tray pattern A, Fig. 22). The contents of the trays were weighed to obtain the weight distribution Wt(x) along the tunnel floor. Then, the contents were irradiated with neutrons, and subsequently, radiation counts were made to deterndne the percentage of ironore. in the trays. The iron ore distribution on the floor, Wfe(X), could thus be calculated. Three normalized cumulative functions were eval-uated to help in the analytical interpretation of the results.

(1) The normalized cumulative total weight distribution along the tunnel floor x

I

wt(x) o Cwt(x)

=

-W t

where Wt is the total weight of the initial sample,

(2) The normalized cumulative iron ore weight distribution on the tunnel floor x

I

wfe(x) cwfe(x) = _0 _ _ _ W fe

where Wfe is the total weight of iron in the ini tial sample, The cumulative fraction of iron ore up to position x

x

The values f'or the above equati~n~ are ShOWIl in Fig. 41. For example, if' a. single cut point isplaced at a distanee 20 cm f'rom the f'eeding slot, then

48%

of' the material processed would·be separated, consisting of'

88%

iron ore and containing ·71% of'. the iron ore in the original sample .. That is, the iron ore-silica mixt~e has been upgradedf'rom

60%

to

88%

iron ore while recovering 71% of' the iron ore. If' f'urther processing by screening were performed then both the purityof' the material and the recovery f'actor

could be in creased.

The second experiment consisted of passing the material as received through tunnel Model 001 with Feeder - 2 with a 70 mesh screen (200 ~) in position. The results Qbtained with the use ·af radioactivity analysis (see Fig.

42)

show that ex cept in the first tray, the iron ore content remains the same regardless of where the cut_~oint is placed. However, since t~e

silica partieles are for the most paPt much larger than the iron ore particles , analysis with the use of sieves on the conbents of four of the trays was also performed yielding the results ShOWIl in Table 11. The dis-crepancy in the fraction of' iron. ore in the infrasized samples as measured by the radioactive and sieving analyses can be attributed to biasing of' the samples since the two analyses were perf'ormed on two mutually exclusive f'ractions of' the contents of' the trays. Although the results may not be conclusive, the fOllowing observations can be noted.f'or the results obtained via the "inf'rasizing followed by sieving analysis" process: . (a) 65% of the total iron ore could be recovered in a mixture eontaining 95% or greater purity iron ore, or (b)

85%

of' the total iron ore could be recovered in a mixture of about

80%

iron ore content. From the point of' view of upgrading the purity of' the iron ore samples, these results seem to be marginally better than those obtained by the first experiment, i.e., via the "sieving followed by the infrasizing" process.

5 . CONCLUSIONS

1. The tunnels designed were simple to construct and compact, and by caref'ul· . screen package design the turbulence can be reduced to less than 1% in the main central portion of' the test chamber.

2. A high degree of uniformity (variations

<

1%) in the main central region was .not achieved with the designs investigated.

3.

The initial conditions upon entering the tunnel have a great influence on the particles' trajectories. To obtain and predict precise classif'i-cation, the maintenance of' uniform initial velocitiesf'or all particles

\

of' IJ:;he same size is of param.ount importance in feeder design.

4.

It has been shown that f'or particles greater than 41 micrometers the classif'ication achieved with the existing ~paratus is quite good.

5.

For particles less than 41 micrometers the inter-particle i~teraction

becomes prominent and the quality of classification was not as good. Special care should be taken in dealing with small particles and very slow feeding rat~ should be used. Tunnel changes that could help, in this area would be a smaller drop height, the elimination of the collec-tion trays to avoid tripping the flow, and provisions f'or humidity control.

---~---

-6.

Horizontal e1utriation in conjunction with screening may be an effective way of performing density c1assifieation.