Professional Documents
Culture Documents
ABSTRACT
Ferrara, G., Preti, U. and Schena, G.D., 1988. Modelling of screening operations. Int. J. Miner.
Process., 22: 193-222.
This paper reviews the Ferrara-Preti screening model and shows that recent laboratory and
pilot plant data validate this approach. The model successfully predicts the partition curve (se-
lectivity function) and therefore the product size distributions, given the feed size distribution.
Some improvement in the application of the model, concerning numerical solution of the equa-
tions and parameter estimation, are indicated.
Information is then given on the influence of some operating conditions on the screen perform-
ance studied through the modification of the model parameters, a way that could be useful for a
more complete approach to modelling of screening operations.
Some applications of the model to improvement of screening operations, sizing of screens, sim-
ulation of screening and of integrated plant operations are discussed.
INTRODUCTION
The screening model considered in this paper has been proposed by Ferrara
and Preti (1975) and subsequently reviewed by t h e m and by other workers
(De Pretis et al., 1977; Schena, 1982; Kelly and Spottiswood, 1982; Herbst and
Oblad, 1984 ). Hess (1983) demonstrated that this model is very accurate.
Results of validation tests are reported by some of the above mentioned au-
thors (Ferrara and Preti, 1975; Schena, 1982; Hess, 1983). Later, some of the
main assumptions on which the model is based were checked by the authors
and the results are reported below.
The model is chiefly aimed at giving a relationship between: (A) variables
which define the mineral fed to the screen ( size distribution); (B) parameters
which characterize the process; (C) dimensions of the screen; and (D) results
of the screening operations. These results can be either in the form of a parti-
tion curve (selectivity function) - - here expressed as screen oversize efficiency
- or in the form of size distribution curves for the products. From these results
-
MATHEMATICALMODEL OF SCREENING
Preti, 1975 ). For the sake of clarity, for both of the conditions the model can
be considered composed of two parts: ( a ) one expressing the kinetics of the
process, through the space- (or time- ) dependent mass balance applied to
populations of particles with different size; (b) another expressing the ratio
between the passage probabilities for particles of different size depending on
the relative size of particle and aperture.
CROWDED SCREENING
Crowded screening occurs when the flow rate is above a critical level Wc
such that the material forms a bed so thick that only particles immediately in
contact with the screen are capable of passage. As long as the upper layers are
able to replenish the contact layer, the mass flow rate of passage remains
constant.
If w (X,1) d X is the mass flow rate per unit width for particles of size from X
to X + dX entering the element d/, the mass balance relative to this element
for the same size class is as represented in Fig. 1. (The definition of the symbols
is given in Appendix I. )
Since it has been shown experimentally that in crowded conditions the mass
of passing particles of size X is proportional to a constant which depends on X
and to the concentration of particles with size X, we have:
aw(X,l)
- - - -k( X) .f( X,l) (1)
0!
where f(X,l) is the concentration of particles of size X, which is given by the
~'w (X,~) dX
density function of the size distribution of the material at the point l, f(x,l),
for x-- X:
w(X,l) w(X,l)
f(X,l) -~o - -- (2)
w(1)
f w(x,l)dx
0
E(X,L)=exp[-k(X)fwd~l(l)l (4)
0
For this approach to be valid, some assumptions are necessary: perfect mix-
ing on the screen along the vertical; plug flow of the material along the screen.
The first assumption has been discussed by Ferrara and Preti (1975). Herbst
and Oblad (1984) have derived general equations which take into account the
dispersion but cannot be solved analytically: neglecting the dispersion an equa-
tion is found having a form similar to the one that we derived.
Eq. 4 can be used to calculate the efficiency curve once the function fd//W(l)
is known or to fit experimental data. For using the model, however, it is pref-
erable to write the equation in a different form, in which some important and
known parameters are explicit. If we consider a second generic value x of the
variable other than X, the ratio of the logarithms of the screen oversize effi-
ciency (eq. 4) for the two variables will define a new variable:
In E(x,l) k(x)
X(x) - I n E(X,I) - k ( X ) (5)
Substituting eq. 2 into eq. 1 and using the expressions 3 and 5 yields:
Integrating the differential eq. 7 between 0 and L and introducing the feed
flow rate per unit width Wo and the density function of the grain size distri-
bution in the feed f(x,0) yields:
D
Eq. 8 is valid for grains of size X between 0 and D; for X>~ D, it will be:
k(X) = 0 E(X,L)=I
Probabilityofpassage
Two functions appear in eq. 8, X (x) and k(X) which can be expressed by a
model derived from the simplest of those proposed by Gaudin (1939) :
p(x) = (D-x)2/(D+B) 2
where D is the mesh aperture (square) and B the wire diameter. Since (Brer-
eton and Dymot, 1973; Ferrara and Preti, 1975) :
k(X) =Wcn'p(X) (9)
where Wc is the critical flow rate and n the number of particle presentations
per unit length, both of them constant under given operating conditions, then
the function X(x) can be expressed as follows:
Z(x) =k(x)/k(X)=p(x)/p(X)= [(D-x)/(D-X) ]~ (10)
According to Gaudin's model, a = 2 for square meshes and a = 1 for wedge
wire screens. Tests have shown that eq. 10 fits very well the experimental data
for a values depending on the vibration conditions of the screen and on the
type of screening surface, which must be determined experimentally.
198
SEPARATE SCREENING
Separate screening occurs when the particles on the screen do not interact
with one another. The differential equation for screening is:
Ow(X,l)
- -s(X) "w(X,l) (14)
al
where s (X) is the kinetic constant for the size X.
The term s (X) in eq. 14 can be derived theoretically by applying the prob-
ability theory as a function of the passage probability p ( X ) and of the number
n of attempts made by a particle per unit length of the screen, according to the
expression:
s(X) =n-ln [1- p ( X ) ] (15)
199
Integrating eq. 14 between 0 and L and introducing the screen oversize ef-
ficiency E (X,L) yields the screening equation for the separate state:
E( X,L ) =exp [ - s ( X) .L ] (16)
for X between 0 and D, while for X >~D one always has s (X) = 0 and E ( X,L ) = 1.
At this point it is necessary to find an expression for the s(X) that has to
be introduced into eq. 16.
One possibility is to utilize eq. 15, replacingp (X) with the Gaudin's expres-
sion corrected by the introduction of a as had already been done for crowded
conditions:
where ]Cois the fraction open area of the screening surface for square mesh. Eq.
16 in this case becomes:
E( X,L ) = e x p { - n ' l n [ 1 - f ~/2(1-X/D ) ~'] "L} (18)
Eq. 18 is applicable only when separate conditions are known to occur and,
given n, [o, ~ and L, it allows the calculation of E(X,L); vice versa, given
E (X,L), [o and L, the equation permits the calculation of the two parameters
n and ft.
Alternatively, and particularly for the study of mixed (crowded and sepa-
rate) conditions, it is possible to use a different form of eq. 18, which though
not exact, is sufficiently approximated for X/D > 0.5. This is the field which is
practically of interest, as all the grains with small X/D pass in the first part of
the screen and only grains with X/D> 0.5 reach the transition zone from
crowded to separate conditions.
Considering that the function R --p ( X ) / I n (1 - p (X)) differs from unity by
TABLE I
less than 10% for X / D > 0.5 (see Table I, calculated for a -- 2 ), in eq. 15 we can
substitute p (X) for In (1 - p (X)) thereby obtaining:
s( X ) =nfg/2" ( 1 - X / D ) ~ (19)
Through the ratio of s (X) to s (x) for x = D/2 ( named Sso) we obtain:
s( X ) =Sso 2~ . ( 1 - X / D ) ~ (20)
which is identical with eq. 11 found for crowded conditions and allows the
influence of the open area to be included in Sso.Eq. 20 has already been reported
by De Pretis et al. (1977), but its derivation has been omitted as well as the
specification that, unlike eq. 11, this is an approximate expression. Sso and a
are functions of the screening surface and of the vibration conditions of the
screen.
Substituting eq. 20 into eq. 16 leads to the following approximate equation
governing the separate state:
MIXED CONDITIONS
If we designate with Lc the critical distance from the feed point for the tran-
sition from crowded to separate conditions and with Wc the corresponding flow
rate, eq. 12 with L=L will p r o v i d e t h e results for screening in the crowded
state; the corresponding screen oversize efficiency is E (X, Lc). In the next por-
tion of the screen, ( L - L c ) , separate conditions will occur and the global screen
oversize efficiency will be:
E ( X , L ) =E(X,Lc) . E [ X , ( L - L ~ ) ] (22)
where E [X, ( L - L c ) ] can be calculated by the use of eqs. 18 or 21.
By using eq. 21 it is, however, possible to obtain a single equation for the
screen oversize efficiency which is valid for crowded, separate or mixed con-
ditions and therefore very useful for characterizing the screening process with
few parameters.
The derivation had already been made by Hess (1983) and is shown again
in the present paper.
Substituting the expressions derived from eqs. 4 and 21 into eq. 22 yields:
E ( X,L ) __exp~_kso.2~ ( I _ X ~ dl o[
( \ uja
0
Assuming:
kso = SsoC (24)
201
where C is a constant with dimensions g/sec cm, substituting into eq. 23 gives:
E ( X , L ) =exp [ -kso "nd'2 " ( 1 - X / D ) ~ ] (25)
where the term
Lc
nd =- | d l / W ( 1 ) + ( L - L c ) / C (26)
o
V A L I D A T I O N OF T H E M O D E L
The basic equations expressing the screening kinetics and the model as a
whole were submitted to checking.
The basic hypotheses underlying the model for crowded conditions and that
for separate conditions were first checked by screening tests carried out at
decreasing flow rates of particles close to 0.5 D in size. A plot of In (AW/AL)
versus In W (Fig. 2a) shows the occurrence of crowded conditions in the region
with constant AW/AL up to the value We; at low flow rates, separate conditions
start to occur, in which zIW/AL decreases proportionally with W. It is worth
noting that the screening conditions change sharply at We, demonstrating that
an intermediate transition zone does not exist. The plot permits the calcula-
tion of ki and si (this latter for W - - l ) . The tests were repeated with a feed
consisting of the same grains with oversize particles in different ranges of per-
centage ( Fig. 2b).
Other checking tests, already reported by Ferrara and Preti (1975), were
202
10
~,w
Z~L
{a)
/,
1
g/sec.cm~
1041 - ~ S i
Screen aperture E] 4 mm
I0-; ~ Porficles size xi ~ 2 mm
~ Open area .56,9 %
~ Amplitude 4 mm
10-: ~ ' Frequency 1200 r p m
7 wc
IO-L
10 I i I I I i I
10-3 10 - 2 10 -? 1 10 10 2 10 3 10 4 105
W glsec.cm
10
A_~w
AL (b)
1
glsec.cm2
10-~
~--- OVERSIZE
10-2
25%
+25%- 50%
lO-3 +50%- 70%
*70% 90%-
10 -4
10-10_3 I I /- I I I I
10 - 2 10 -1 1 10 10 2 10 3 10 4 10 5
W glsec cm
Fig. 2. Screening tests showing the transition from crowded to separate conditions. ( a )
Feed=particles having xi=2 m m ( size class - 2 . 3 8 to 1.70 ). ( b ) Feed consisting of the same
particles as ( a ) + oversize in different ranges of percentages.
ditions. The alignment of the experimental points along straight lines has al-
203
Io, Iu.
Fig. 3. Scheme of laboratory screen divided in four sections.
3
1
o
0
-1
I
in klX~
k~o -2
i
-3
-4
-Sl
-6 _ _ 1 i i i
0.5 1 1.5 2 2.5
-in (1-XIO)
lowed the verification of eqs. 4 and 16 for the two screening conditions. The
tests were carried out through a laboratory screen divided into 4 sections, by
screening crushed limestone containing all the classes in the range 5 to 0.2 mm
(aperture of the screen 4 m m ) and collecting the underflow separately (Fig.
3 ). Deviations from the crowded condition model were observed when the bed
was too high on the screening surface, particularly as regards the fine classes,
probably because the assumed perfect mixing along the vertical had not
occurred.
The relationship defining k ( X ) as a function of X / D (eq. 11) was verified
by plotting In [ k (X)/kso ] versus In ( 1 - X / D ) . The straight lines obtained had
slope values, a, varying around 2 (Gaudin's law) and passing through the point
of coordinates k (X)/kso = 1 and (1 - X / D ) = 0.5. Fig. 4 shows the experimen-
tal points for one of the tests, as example, and the straight lines corresponding
to ~r values from 1 to 3, a range of practical interest.
The model as a whole was verified on the basis of the results of screening
tests performed on the screen of our laboratory and on the industrial-size
screens of the pilot plants of S K E G A A.B. and SVEDALA A R B R A A.B. ( size
of the screens: 1.133.60 and 1 . 0 0 3 . 5 0 ) . In these pilot plants the screen
underfiow is collected separately for a number (8 or less combining the un-
derflows ) of screen sections as illustrated in the scheme of Fig. 3. This figure
shows a subdivision into 4 sections, as is the case for our laboratory screen; it
can be seen that, when the weights and size distributions of U1, U2,/-/3, U4 and
204
W4 are known, it is possible to know Wo, W1, W2, W3 and, hence, the screen
oversize efficiency curves for each section.
For different operating conditions, type of vibration and of screening sur-
face, kso and a were calculated for successive sections and/or for increased
TABLE II
Pilot plant screening test: free swinging circular motion; frequency 1100 rpm; amplitude 7 mm;
screen slope 15; wire cloth; 4 mm opening; open area 55%; feed rate 59.5 t/h m
TABLE III
Pilot plant screening test: free swinging circular motion; frequency 1100 rpm; amplitude 7 ram;
screen slope 15 ; wire cloth; 8 mm opening; open area 67%; feed rate 98.6 t / h m
TABLE IV
Pilot plant screening test: free swinging linear motion; frequency 700 rpm; amplitude 18 ram;
screen horizontal; wire cloth; 4 mm opening; open area 55%; feed rate 65.7 t / h m
length values of the screen; the values obtained are constant with a good ap-
proximation, provided that the operation is carried out in crowded conditions
but without overloading the screen.
Tables II, III and IV give, for three pilot plant tests, the rough and final
values (calculated with the method explained in the next section) of kso and
for increased lengths of the screen. They show the order of magnitude of the
observed deviations from a constant value, that are chiefly due to experimental
errors and probably, sometimes, to vibrating condition of the screen that are
not completely constant on the whole surface. In the last column the mean
standardized fitting errors are reported.
Tables II and III refer to screening in the same vibration conditions with
circular motion, but different screen apertures (4 m m and 8 m m ) . Tables II
and IV refer to the same screen aperture ( 4 m m ) , but with different vibrating
conditions (circular and rectilinear vibration). The results show that the cir-
cular motion gives higher a values and lower kso values than the rectilinear
vibration. Therefore the first condition is more appropriate for a better sepa-
ration of the fines, and the second for a better separation of near-submesh
particles.
Fig. 5 shows, for the test of Table II, the simulated efficiency curves and the
experimental points considering three screen lengths.
Looking through a number of data the following observations were made.
SCREEN LENGTH
SQUARE OPENING
59,5 Ilh m
>-
u.J
u,..
lc D
I
0!
0 1 2 3
PARTICLE SIZE. mm
Fig. 5. Pilot plant screening tests.Simulated efficiencycurves and experimental points for three
screen lengths.
206
TABLE V
Pilot plant screening test: free swinging linear motion; frequency 746 rpm; amplitude 16 ram;
screen slope 5 ; rubber deck; rectangular opening 6 20 ram; open area 26%; feed rate 71.2 t/h m
(a) Sometimes for the finer size classes is observed a screen undersize effi-
ciency [1-E(x,L)] lower than that for the size class immediately coarser.
This fact, simply due to adhesion of very fine particles to the coarser ones, has
nothing to do with the true screening and the model.
If the cause of anomaly is not eliminated (e.g. moisture) or the results are
not corrected, the fitting accuracy will suffer.
For example, for the test of Table V (Fig. 8, curve for 71.2 t / h m) where the
anomaly occurs, higher values of the standardized fitting errors were observed.
For the same reason, probably, the experimental points for class 0-1 mm in
Fig. 5 are slightly lower than predicted by the model.
(b) Some deviation of simulated from experimental results can be observed
in Fig. 5 for the near-submesh class 3-4 mm.
This may be due to different reasons, including the large sieve intervals. In
particular, it can be observed that the model is very sensitive to the value of D
and therefore using in simulation a value De, slightly different from the nom-
inal one (lower in case of Fig. 5), the fitting errors could be reduced.
The subject will also be discussed in a later section for rectangular openings,
where the above reported suggestion becomes an imperative need.
On the whole, considering also the source of inaccuracy connected to the
large sieve intervals used for the tests of Tables II, III and IV (e.g.,
0 - 1 - 2 - 3 - 4 - 5 - 7 m m for Table II), the model proves to be reliable and can easily
be made to simulate actual screening performance by proper selection of the
parameters.
The model was also verified by Hess (1983), using eq. 25 and fitting the two
parameters ndk~oand a. The comparison of experimental and simulated data
has shown the lowest mean standardized fitting error of all the models he
considered.
207
The use of continuous functions for the definition of the model, such as were
employed in the present study, is more general and, in principle, improves the
model accuracy, but requires the knowledge of continuous functions defining
the size composition of the feed.
However, the size composition of the feed is generally known for discrete
values of particle size; in this case, eq. 12 for the crowded condition and eq. 21
for the separate condition must be written in discretized form.
This transformation leads to the following equations, already reported in
previous papers by the authors:
Wo~ ~ y j o l [ E ( x i , L ) Z J i - 1 ] + l n E ( x i , L )
(j=l Zji
~ Yjo}
i=n+l
E(xi,L)=l forn+l<~i<m
where the index i refers to the size class whose behaviour during screening is
being studied, and j refers to the other size classes present in the feed which
affect the behaviour of size class i; Xji = [(D - xi) / (D - xi ) ] ~. The classes from
i to n are smaller than the mesh aperture of the screen, while the classes from
n + 1 to m are larger.
For separate conditions:
E(xi,L) = e x p [ - s ~ o 2 ( 1 - x J D ) ' L ] for l<~i<.n
E(xi,L) =1 for n+ l <.i<m (28)
Not taking into account the separate conditions for which the solution is
simple and independent of the grain size composition of the feed, attention is
focused on the method for the solution of eq. 27. This equation, in order to
obtain E (xi,L) for a given value of L, has to be solved for the n classes having
xi<D.
The numerical solution is easier by the following variable transformation:
E(xi,L) = e x p ( ? i ) (29)
Substituting into eq. 27 we obtain the following equation in the variable 7:
n
V(ri)=zYi[exp(~iZji)-l]+ ~ yjoyi+k~o2
o (\ 1 - ~ xi )o L
~oo=0 (30)
j = 1Xji j = n + 1
Eq. 30 can be easily derived and solved using the methods of numerical analysis
208
( Newton's method). The values for the screen oversize efficiency are obtained
from eq. 29 by substituting into it the roots of eq. 30.
CALCULATION OF M O D E L P A R A M E T E R S
If only the fitting of the parameters defining the curve of the screen oversize
efficiency is of interest, regardless of the conditions present on the screen (eq.
25), it is sufficient to determine ndkso and a. Conversely, if eq. 27 is to be used
for design and simulation purposes then kso and a must be determined pro-
vided that the screening occurs in crowded conditions. In this case the above
procedure cannot be followed because nd is unknown (eq. 26). A procedure is
proposed which can be applied when the screening results are known for one
screen section or, better still, for more screen sections, so as to permit a more
accurate estimation of the parameters. The procedure is also applicable when
the screening results for the whole screen are known, provided that on the
whole screen crowded conditions occur. This procedure consists of the follow-
ing two steps.
(1) Calculating a rough estimate of kso and a by means of eq. 4, re-written
in the following form for the generic class i and for the screen length from Lk
to Lk+ 1:
Lk+l
l n E i = - k i | dl/W(l) (31)
Lk
When test results are available for screening sections (Fig. 3 ) at Lk, Lh+ 1, then
Wk and Wh+ 1 are known and hence a rough estimate of the integral of eq. 31
can be obtained by considering W(l) constant and equal to ( Wh+ Wk+l)/2,
so that the value for the integral is 2L/( Wk+ Wh+l). In this way, ki can be
calculated from eq. 31 for each class i and kso and a can be obtained from a set
of equations for which the generic one, obtained from eq. 11, is:
Inki = lnkso + a In [2 (1 - xi/D) ] ( 32 )
which we prefer to solve using the method that minimizes the absolute deviation.
(2) Once a rough estimate of kso and a has been obtained, these values are
used as guess values for a more accurate calculation method. This consists in
simulating the process by using eq. 27 for each class i and calculating the
function:
where Ei and Ei* are the experimental and simulated values of the screen over-
size efficiency, respectively, and zi are appropriate weighting coefficients.
209
f f(X,O) . E ( X , I ) . d X
E( xi,l) - xi xi+l (36)
f f(X,O).dX
xi
(1) A correct procedure would require the estimation of the true values of
the model parameters, kso and a, i.e. the values corresponding to the true screen
oversize efficiency. These values only, in fact, are independent of sieve inter-
vals and feed size distribution. To this end, in data processing, the experimen-
tal values of the screen oversize efficiency (related to wide sieve intervals)
have to be compared with the simulated ones, calculated using narrow intervals
( or preferably the continuous function) and then averaged on the wide inter-
vals corresponding to the experimental data by eq. 36. Therefore, the same
procedure explained in the preceding section on calculation of model parame-
ters can be applied, including the minimizing method of eq. 33, being Ei* av-
eraged screen oversize efficiency values as explained before.
(2) Using the true values of the model parameters, the screening results
corresponding to different sieve series (e.g. 1/2, l/v/2, 1/~/2 or an irregular
series) can be simulated.
(3) Conversely, if the available values of model parameters correspond to a
discrete system (e.g. to a particular testing sieve series), only screening results
within the same discrete system can be correctly simulated. However, if this
system is based on sieve intervals narrow enough, the model parameters values
can be considered a good approximation of the true ones.
The kinetic parameter kso is the mass flow rate per unit area for particles of
size X-- 0.5 D, assuming the feed made up of grains all of the same size and the
screening occurring in crowded condition. This parameter can be measured in
g s -1 cm -2 or i n t h -1 m -2 (1 g s -1 c m - 2 = 3 6 t h -1 m - 2 ) .
The significance of kso is similar to that of the so-called empirical or basic
capacity in the well known formulae for screen sizing ( Colman, 1978; Nichols,
1982).
The parameter kso depends chiefly on the mesh aperture, as does the capacity
mentioned above. It also depends on parameters characterizing the screening
surface (fraction open area, aperture shape, type of screening surface ), on the
vibration characteristics of the screen (frequency, amplitude, form of the os-
cillation) and on the inclination of the screen.
The parameter a affects the ratio of passage probabilities for particles of
different size (eq. 10) and, once kso is known, allows k(X) to be calculated for
the generic size X by means of eq. 11. To illustrate further the meaning of a,
Fig. 6 gives a plot of k(X) versus X/D, assuming kso= 1. As can be seen from
the plot, for high a values the k(X) values for the fine classes are high,
( X/D < 0.5 ), while those for the near-submesh particles are low (X/D > 0.5 ).
Conversely, for low a values the near-submesh particles remarkably increase
their k(X) value.
211
0 O.S
D
9C \ 2
. . . . 0" = I
'\\ o'=2
............. O" = 3
7c
\
>.- \
Z
\
\
u_ sc
\
LLI
\
\ \
w z,.( \
N
\
Z
3o \
, \ \
',, \ \\
",,
i t i
o 1 2 ; ~
PARTICLE $1Z, mm
Through the model eqs. 12 and 27, the parameters kso and a completely
define the screening process in crowded condition. Thus, the influence of the
operating conditions on the screen performance can be investigated much more
easily through the study of their influence on kso and a. It then becomes need-
less to investigate the effects of the feed size, the flow rate and the screen
dimensions (length and width), since these quantities are taken into account
- - either explicitly or implicitly m in the equations of the model.
The following effects still to be investigated are then: (1) parameters char-
acterizing the screen surface (aperture shape, fraction open area, type of
screening surface material); (2) vibration characteristics of the screen (fre-
quency, amplitude, form of the oscillation, inclination of vibration, whether
rectilinear or elliptic) ; and ( 3 ) inclination of the screening surface, also in
connection with the vibration characteristics.
The influence of the above listed parameters on kso and ~ was investigated
only partially. The main results obtained are reported in the following sections.
APERTURESHAPE
The shape of the aperture affects both kso and a. In passing from square to
round apertures, k~odecreases as if the round aperture corresponded to a smaller
square aperture; a on the other hand, increases, indicating that in a relative
sense the fine particles pass much more readily than the coarse ones and that
the conditions occurring are most unfavorable for the passage of near-submesh
particles.
In passing from square to elongated rectangular apertures, and assuming D
as the shorter side, k~o should increase and a decrease. Near-submesh and also
elongated particles pass into the underflow readily; in the test sieving, these
particles are classified as oversized with respect to the mesh aperture D. In
214
SCREEN LENGTH
RECTANGULAR OPENING
1
I. 2oo-
>-
71,2 tlhm
132,2 t/h.m
\
0 1 2 3 $,. S 6 7 B
PARTICLE SIZE, mm
Fig. 8. Pilot plant screening tests. Simulated efficiencycurves and experimental points for two
feed rates and different screen lengths. Rectangular openings. Tests of Tables V and VI.
Fig. 8 shows the simulated efficiency curves and the experimental points for
two tests carried out under the same conditions except for the feed rate ( Ta-
bles V and VI ). For the higher feed rate 4 screen lengths have been considered.
It can be observed that, in spite of the rectangular shape of the aperture, the
equation of the model fits the experimental points very well and t h a t the stan-
dard errors are of the same order of magnitude found in the previous applica-
tion for square meshes.
The conveying speed of the material on the screen, the trajectory of the
particles, the n u m b e r of attempts per unit length made by a particle to pass
through the mesh and the passage probability m and hence also the value of
the p a r a m e t e r kso - - depend on the frequency, amplitude and shape of the
oscillations, on the inclination of rectilinear vibration or of the major axis of
elliptical vibration and also on the inclination of the screening surface. The
ratio of passage probabilities for particles of different size and hence the value
of a, are also dependent on the above parameters.
The optimization of the screen performance can be attained by maximizing
kso thereby ensuring the highest capacity of the screen; for a, high values will
215
+
rpm e~
1000 50
1000 40
1000 50
1200 50
1200 4 0a
1200 ~0
Jr 1400 30
1400 40 e
1400 50
REPEATEO TESTS
I'(5o I 1000 50
0
X
4-
i i s i i i i
1 2 3 4
Kv
l D
x
0
o ,7
Q x
0 I I I I | r I i
1 2 3 L, 5
Kv
Fig. 9. Influence of Kv on kso and a. Horizontal screen, rectilinear vibration, square mesh (4 mm
aperture).
217
In any case, it is believed that the results obtained are interesting in that
they show that a more accurate choice of the vibration conditions of the screen
allows wider possibilities for the optimization of the screen performance. In
fact, Fig. 9 suggests that operating with Kv at about its optimum permits the
ks0 value - - and hence the capacity of the screen - - to be nearly doubled.
However, since the same value of Kv can be obtained with different values
of amplitude, frequency and angle 01, a further search for optimum conditions
is possible on the basis of these variables.
It is also possible that for values of the variables beyond the range investi-
gated by the authors the screening conditions could be unsatisfactory despite
the Kv value. A study in this regard would be necessary.
Some information on the effect of the individual variables taken from stud-
ies made previously or still under way, is reported below.
With increasing amplitude, kso decreases, whereas a increases markedly (De
Pretis et al., 1977 ). A similar effect is obtained by increasing the frequency.
An increase in the inclination 01 of the rectilinear vibration results in an
increase, followed by a decrease, in both k~0 and a; their maxima, however,
correspond to different angles (Ferrara and Preti, 1975).
For counterrotating circular motion, kso decreases slightly, while a decreases
considerably (e.g. from 2.5 to 1.5) as compared to the opposite rotating con-
ditions. This type of screen motion is therefore particularly favorable for the
passage of near-submesh particles (De Pretis et al., 1977), even if the screen
capacity is reduced.
1.6 /
80 1.8 /+///
UNDERFLOW OVERFLOW /
% RETAINED 2 % PASSING //
60
~0
if/i/Ill~'~
o 6= 1
\
20
-----.,c.----- 0 = 2.6
! 2 3 ~ 5 6 ? 8
PARTICLE SIZE, rnm
Fig. 10. Simulated screen results for different conditions of vibration. In the center are reported
the undersize efficiency curves for different a values, on the sides the size distributions of the
products for maximum and minimum a.
using smaller screens and screening surfaces with wider apertures and, hence,
less subject to blinding and more resistant to wear.
Fig. 10 illustrates an example of screening at 4.75 mm simulated for different
conditions of vibration which affect the a value. The evolution of the screen
undersize efficiency curves for a values from 1 to 2.6 and the size distribution
of the products for the two e x t r e m e values are reported, showing a quite re-
markable difference in cut points and product specifications. The cut point
can be further shifted by repeating the operation with multideck screens.
The fine screening achieved with wide apertures is now of considerable in-
terest in that it allows the solution of problems associated with blinding of
screening surfaces, screening of moist materials, poor screening performance
with small apertures, cost and wear of the screen surface. Some types of screens
having the above mentioned advantages have recently found widespread ap-
plication in the removal of fines from raw coal without the need of wet screening.
Other applications of the screening model can be of interest to the operators
when the performances of an existing plant are not as required because of
insufficiency or inefficiency of the screens. These screens may become over-
loaded either because they are undersized or because of insufficiency of the
other machines (e.g. secondary crushers), which do not yield a sufficient pro-
219
The model was initially intended as a new approach to the sizing of screens
(De Pretis et al., 1977). To this end, however, it is necessary to know a series
of k~o reference data for different apertures of the screening surface, that is, a
diagram of the type giving the empirical or basic capacity for the sizing meth-
ods proposed by the manufacturers. Moreover it is necessary to know the vari-
ation of kso around the reference value and of a as a function of all the
parameters on which they are dependent. As mentioned earlier in this paper,
neither the grain size composition nor the flow rate of the feed appear among
these parameters but only characteristics of the screening surface, vibration
characteristics and the inclination of the screen.
At present, only a limited amount of data is available to enable us to assign
values to kso and a. A joint effort with screen and screening surface manufac-
turers would be needed in order to produce tables or diagrams from which the
values of kso and a can be derived for a number of selected cases.
If, however, from tests performed on screens installed in other plants or on
pilot plants of the manufacturers the values of k~o and a can be known regard-
less of the materials screened and of their size composition, the plant results
obtainable with the same types of screens and vibration conditions can be sim-
ulated with accuracy except for those cases where the grain shape in the two
materials differs markedly.
By operating in this way the sizing can be more accurate than when using
classical design methods.
The use of the model for simulating the results may be very helpful in the
checking of the dimensions adopted for the machines whatever calculation
method is applied in designing the screen. The effect of the various external
factors on the screen performance (e.g. variations in the feed ore characteris-
tics, or in the behavior of other machines included in the circuits) can be in-
vestigated accurately by sensitivity and risk analysis as already reported for
other problems in a previous paper by Ferrara et al. (1984).
This design and checking procedure is particularly proposed for application
to the study of plants in which screening is of remarkable importance for the
efficiency of the process, energy saving and for obtaining products with closer
specifications. In these cases a procedure enabling a more accurate prediction
of the results is to be preferred.
In this connection it is to be noted that the calculation methods for screens,
220
even if conservative, not always allow one to avoid possible failures. Difficult
screening conditions exist (e.g. with circulating load containing a high pro-
portion of near-submesh particles) for which calculation methods may not
obtain adequate results. Especially when such conditions are to be expected,
the use of simulation procedures to verify the plant efficiency is necessary.
ACKNOWLEDGEMENT
This study was carried out with the financial assistance of the Ministry of
Education (M.P.I.). The authors wish to thank the companies SVEDALA-
ARBRA A.B. and SKEGA A.B. for their cooperation with regard to pilot plant
screening data. In addition,,thanks are due to the graduate mining engineers
F. Nassivera and S. Pinat for their collaboration in performing tests and data
processing.
A P P E N D I X I - - D E F I N I T I O N S OF S Y M B O L S
REFERENCES
Batterham, R.J., Weller, K.R., Norgate, T.E. and Birkett, C.J., 1980. Screen performance and
modelling with special reference to iron ore crushing plants. Europ. Symp. Particle Technol.,
Amsterdam.
Box, M.J., 1965. A new method of constrained optimization and a comparison with other methods.
Comp. J., 8: 42-52.
Brereton, T. and Dymott, K.R., 1973. Some factors which influence screen performance. Proc.
10th IMPC, London.
Colman, K.G., 1978. Selection guidelines for size and type of vibrating screens in ore crushing
plants. In: A.L. Mular and R.B. Bhappu ( Editors ), Mineral Processing Plant Design. AIME,
New York, N.Y. pp. 341-361.
De Pretis, A., Ferrara, G., Guarascio, M. and Preti, U., 1977. A new approach to screening design.
Proc. 12th IMPC, S~o Paulo, Brasil.
Ferrara, G. and Preti, U., 1975. A contribution to screening kinetics. Proc. 1 l t h IMPC, Cagliari,
Italy.
Ferrara, G., Guarascio, M. and Schena, G., 1984. Modelling and simulation of integrated plant
operations of mineral processing. In: J.A. Herbst (Editor), Control '84. AIME, SME/TSM,
New York, N.Y., pp. 153-165.
Gaudin, A.M,, 1939. Principles of Mineral Dressing. Mc Graw-Hill, New York, N.Y.
Herbst, J.A. and Oblad, A.E., 1984. A population balance model for screening. Proc. 9th Powder
in Bulk Solids Conf., Chicago.
Herbst, J.A., Schena, G.D. and Fu, L.S., 1986. Computer aided design of comminution circuits.
SME-AIME Meeting, New Orleans, La.
Hess, F., 1983. Mathematical Modelling of Screen and Related Units for Plant Simulation. Ph.
D. Thesis, University of Queensland, Australia.
222
Kelly, E.G. and Spottiswood, D.J., 1982. Introduction to Mineral Processing. John Wiley and
Sons, New York, N.Y., pp. 185-189.
King, E.H., 1977. How to Determine Plant Screening Requirements. Chem. Eng. Progr., 73 {5):
74-79.
Kluge, W., 1951. Neuzeitliche Siebmaschinen fuer die Aufbereitung. Erdoel Kohle, 11: 705-711.
Nichols, J.P., 1982. Selection and sizing of screens. In: A.L.Mular and G.W. Jergensen (Editors),
Design and Installation of Comminution Circuits. AIME, New York, N.Y., pp. 509-522.
Schena, G.D., 1982. The Processing of Industrial Screening Data. A Modelling Approach. Internal
Report, Ist. Miniere e Geofisica Appl., Universita' di Trieste.
Schranz, H. und Bergholz, W., 1954, Die Bewegungsvorgaenge bei Wurfsieben. Bergbauwissen-
schaften, 8: 223-234.
Stoff, F., 1963. Betrachtungen fiber den Einfluss der Schwingungsdaten auf den Siebvorgang.
Aufbereitungs-Techn., 11: 426-466.
Whiten, W.J., 1972. The simulation of crushing plants with models developed using multiple
spline regression. 10th Int. Syrup. Application of Computer Methods in the Mineral Industry,
(J. S. Aft. Inst. Min. Metall., Johannesburg), pp. 257-264.