Monitoring of Medium Rheology in DM Cyclone Plants
Monitoring of Medium Rheology in DM Cyclone Plants
Monitoring of Medium Rheology in DM Cyclone Plants
T.J. NAPIER-MUNN
T.J. REEVES
J.O. HANSEN
ABSTRACT
nd
2 Samancor Symposium on DM Separation - 1985
2
NOMENCLATURE
A Area
Cv Concentration of solids by volume
a,b,c Constants
d Particle size
F Force
f(..) Function of ..
H Head (m of slurry)
h Height of coil zone
K1,2 Constants
k Factor in sedimentation equation (1.1)
n Exponent
Rei Cyclone Inlet Reynolds Number
S Shear rate
s Velocity
vs Interface sedimentation rate
vso Equivalent Stokesian sedimentation rate (single particles)
x Distance
α, β Exponents
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
3
1. DEFINITIONS
The term "rheology" refers to the flow properties of materials. In the case of
unstable (settling) suspensions, such as dense media, the two rheological
properties of interest are stability and viscosity.
Stability implies the existence of a two-phase system, in which one phase (the
solids) settles or flows at a certain rate through, or relative to, the second phase
(the water). In dense systems (those in which the solids volume concentrations are
high, say > 20%), the solids usually settle under gravity en masse, creating a clear
solid-water interface, which is observed to move at a constant rate, until a
compression zone is reached. Such sedimentation, in which particles of different
sizes settle at the same rate, is termed hindered settling. The stability of a dense
medium is conventionally defined as the reciprocal of the sedimentation rate.
Stability may be defined either in terms of the interface sedimentation rate or the
change in density with respect to time at an arbitrary point in the suspension; the SI
units of stability are therefore
s m-1 or s m3 kg-1
respectively, although in the case of the latter definition we have adopted the
reciprocal unit
gℓ-1 s-1
where k is a multiplying factor (depending on particle size and shape) implying the
existence of water bound to the particles and thus effectively removed from the
suspension.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
4
interpreted as the resistance of the fluid (or suspension) to shear. When a shearing
force is applied to a fluid, and flow is initiated, a velocity gradient is set up in the
fluid as a consequence of its internal friction; the fluid layers can be thought of as
sliding over each other, each layer being retarded by the adjacent layer. The rate of
shear, S , is expressed as a rate of change of velocity with distance from the
shearing force:
ds
S (1.2)
dx
The shearing force acts over a given area, and is therefore expressed as a stress:
F
shear stress (1.3)
A
= f( S ) (1.4)
The simplest fluids are those for which the flow curve is a straight line passing
through the origin:
= S (1.5)
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
5
Pseudoplastic: = K1 ( S )n n < 1
(1.7)
Dilatant: = K1 ( S )n n > 1
Equation 1.7 is termed the power law, which is an empirical relation in which K is a
consistency index (higher for more viscous fluids) and n is an exponent, constant
over large ranges of shear rate. The dimensions of K are not those of viscosity, but
depend upon the value of n, which is dimensionless. [K] = ML-1 Tn-2.
It should be noted that these flow curves are idealised, and real fluids (and
suspensions) often exhibit features of more than one type, over certain shear
ranges. One combinatory type in particular has been found useful by the authors in
dealing with dense media[1]. It incorporates the features of the Bingham plastic and
power law types:
= O + K S n (1.8)
Experimental work has suggested that, for ferrosilicon suspensions, n > 1 (i.e. there
is a tendency to dilatancy at the higher shear rates)[1].
It is very important to note that, for the more realistic models of equations 1.6, 1.7
and 1.8, it is impossible to define a fixed, true viscosity over the whole range of
shear. This has led to the widely used, and widely misunderstood, concept of
apparent viscosity, a, which is usually defined as a local (point) ratio of shear
stress to shear rate:
τi
ai = (1.9)
S i
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
6
where the subscript, i, identifies the point on the flow curve at which the viscosity is
being estimated. The apparent viscosity may be a useful concept in considering the
hydrodynamics of DMS processes, since it is probably the "true" viscosity
experienced by an ore particle moving relative to the medium; in such a case, the
subscript, i, defines the local mean shear rate, which is a function of the particle
size and shape and the ambient particle-medium velocity. However, the local shear
rate, and thus the apparent viscosity, is very difficult to quantify experimentally.
Further discussion of the definitions of viscosity and stability are beyond the scope
of this paper, but a detailed review of the rheological properties of suspensions, and
particularly dense media, can be found elsewhere [1].
Anyone who has operated a dense medium bath or drum will have a good
understanding of the concept of medium consistency or viscosity. When the
medium is like porridge, the separation deteriorates. The reasons for this are of
course well understood, and the explanation is usually given in terms of the
reduction in the terminal velocities of the ore particles, as described by the
hydrodynamic sedimentation laws such as those of Stokes (in laminar flow) and
Newton (in turbulent flow).
The dense medium cyclone was born out of the fortuitous observation that coal
preferentially concentrated in the overflow product from a cyclone being used to
thicken loss medium for regeneration in a static bath washing plant. The midwives,
the Dutch State Mines organisation, exploited the discovery very effectively, and
published widely in the early days of its development. The separation was explained
by Krijgsman in a 1952 paper[12] in terms of a "barrier" of medium particles, which
built up in the lower part of the cyclone. He emphasised the importance of utilising
medium particles of a size suitable to build up the barrier, and this led to the
preferred use of magnetite as the medium for coal preparation.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
7
It was immediately recognised that the media used for many years in static bath
separators were not appropriate to cyclones. This was due to the high centrifugal
forces utilised in the cyclone (2-3 orders of magnitude higher than gravity), which
rendered the traditional media highly unstable. Anyone who has run a 100 mm
cyclone at high pressure with 48D ferrosilicon, as one of the authors has done [13],
will know to his cost the effects of extreme instability - sludge in the underflow, clear
water in the overflow, and a horizontal Tromp curve.
Cyclone operations clearly required a finer, more stable range of media, and the
ferrosilicon manufacturers responded accordingly; to this day, there are grades of
atomised ferrosilicon sold under the names "Cyclone 60" and "Cyclone 40" [13].
Although the gross effects of instability were immediately apparent, the appreciation
of the more subtle effects of medium rheology on cyclone performance has dawned
only slowly, and has yet to be fully recognised in some quarters. Indeed, many of
the early publications on the cyclone cited the absence of viscosity effects (due
supposedly to the high shear forces prevailing) as one of the many advantages of
the device. A diligent search of the early literature, however, reveals a number of
references to effects that are clearly rheology-dependent. Herkenhoff[14], for
example, monitored the change in the separation with different media size
distributions, and also noted that the ore separation changed when the medium was
magnetised (which is known to influence viscosity [13]); his results certainly imply
that 50 (the separating density) reduces as the viscosity increases. Sokaski and
Geer[15] noted changes in performance with different sized media, and Davies et al
[16]
drew attention to the importance of medium stability, as expressed in the
prevailing density differential between underflow and feed medium; a similar
conclusion was drawn in the more recent work of Collins et al [17]. The first workers
to investigate explicitly the influence of medium viscosity on cyclone performance
were Lilge et al [18,19] (experimentally) and Tarjan[20] (theoretically). In both cases,
the apparent viscosities at different radii were inferred from the prevailing (apparent)
shear rate, and then related to the observed density separation. Olfert [21] mentioned
the (rheological) effect of medium contamination upon the separation.
Unfortunately, most workers did not represent their separations in terms of the
Tromp curve, the direct criterion of separation performance, but rather by chemical
assays of the products. However, it can be inferred that practically all the literature
on unstable suspensions concludes that the 50 decreases as viscosity (or stability)
increases, and that the separation quality (usually expressed as the Ep-value) will
deteriorate as the viscosity rises, although the Ep will actually depend very much
upon the prevailing 50, the feed density distribution and other factors (Gottfried[22]
has even demonstrated that one can expect the relation 50 Ep from the
mathematical properties of the generalised partition curve).
The present authors and their colleagues have, since 1972, carried through a
comprehensive academic pilot-plant and production-scale research programme on
the performance of cyclones with ferrosilicon media; some of the results have been
published elsewhere[1,23-26]. The principal conclusion of all this work is that, once the
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
8
geometry of the cyclone has been defined, the properties of the medium are
process-determining. This conclusion has been confirmed many times, sometimes
dramatically, in post mortems on production plant performance.
The research has demonstrated, among other things, that medium size distribution
must be carefully selected and maintained for a particular application, and that an
increase in viscosity caused by non-magnetic contamination reduces the 50;
Figure 2, taken from Hampel's work (reported in Ref. 23), shows the latter trend
clearly. It has also been demonstrated that cyclone performance is a function of
both stability and viscosity independently, and that it is therefore desirable to
monitor both quantities separately.
To illustrate the latter point, it is helpful to consider a little of the theoretical basis of
DM cyclone separations, even though the theory is by no means fully developed.
It can be shown by dimensional analysis[24] that the 50 for a cyclone operating with
a stable (non-settling) medium is given by:
50 ρ f
K Rei - α d β (2.1)
ρf
(where and are exponents whose value depends on the particle flow regime.)
The inlet Reynolds number, Rei, contains a medium viscosity term in the
denominator, and the 50 is thus seen to be a direct function of viscosity. This
relation has been confirmed experimentally[24]. (The work also showed that Ep
increased with viscosity.)
Equation 2.1 is not appropriate for a conventional unstable medium. In this case, it
has been shown that[1,25]:
l 1 8Hz G2
u = l + (f -l) 1 - exp.G1 - (2.3)
R m R m Q f
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
9
Here z is a sedimenation (stability) term, and the pressure head (H) and flowrate
(Qf) terms are related through a viscosity term:
β
ηa
H = K2 Qf (2.4)
ρ f
(where and are exponents whose value depends on the prevailing flow regime.)
Viscosity and stability therefore appear independently in the expression for u, and
thus for 50. (It should be noted that the relation of 50 = u proposed by some
workers [16,17] is a special case of equation 2.2.)
3.1 INTRODUCTION
As noted earlier, it has been found experimentally that the size distribution of
ferrosilicon media is a process-determining variable in cyclone operation.
Accordingly, many plants now monitor the size distribution of their in-circuit medium
as a routine function. However, such analysis is difficult to carry out reliably, and is
tedious and slow. In any event, it is really only a secondary indicator (with other
parameters) of the fundamental process variables, stability and viscosity. The
stability meter was therefore conceived as a faster, more reliable and more relevant
method of monitoring an important medium property in the plant environment.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
10
Mudline descent rate (where visible) depends upon the position of the depth
interval, relative to the base of the vessel, over which timing takes place.
The approach taken at the Diamond Research Laboratory has been to define a
vessel, a 300 mm long tube of 19 mm internal diameter, and to select a zone
centred 55 mm below the upper end. After stirring by end-over-end inversion a
prescribed number of times, the density is monitored and a rate of change of
density is measured. Since the density drops off in a sigmoid curve (Figure 3), and
sometimes with a slight rise at the start of the test, the rate of change of density
over the most linear portion of the curve, found empirically to be between 68 and
32% of the total density change, is measured. This rate of change of density, in
units of g/ℓ/s (to avoid small numbers), is defined as the settling rate, and is an
inverse index of stability. Figure 3 shows a plot of density with time as measured by
a sensor coil situated near the top of a tube containing a ferromagnetic medium.
When the suspended particles are highly ferromagnetic, their presence within a
volume of diamagnetic liquid will change the permeability significantly. A coil
containing this volume will then experience a change in inductance, and this can be
measured by electronic techniques and related back to the volume fraction of
ferromagnetic material and hence the pulp density. This assumes that non-
magnetic solids do not contribute significantly to the measurement. The
development of a suitable size of coil and sample tube, the refinement of the
electronic circuits required to extract a density-equivalent signal, and the method of
calculating a rate from the time-varying signal are all problems which have been
addressed during the development programme. The meters are of digital design,
which provides the following advantages:
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
11
1. Ease of calibration using a calibrating slug provided with the meter. This
operation is carried out on a prompt from the meter display and is a pure
"software" operation: there are no knobs or screws to turn.
2. Accuracy of density and settling rate readings is improved.
3. Results of repeat tests can be averaged internally and the mean displayed.
4. Operation by semi-skilled plant staff is made possible by a prompt menu (in
any of four selectable language options) which details the procedure needed
for calibration and test.
Stirring of the medium prior to the settling phase is accomplished by inverting the
sample tube end over end at a preselected speed. This appears to be effective for
magnetite and milled ferrosilicon at all densities, and for atomised ferrosilicon at
start SGs higher than 2,4. Below this density, the coarser atomised particles are not
properly suspended before the test commences. By integrating the stirrer drive and
readout electronics in a single cabinet of dimensions 500 x 300 x 200 mm, a
compact instrument, which can be used in the plant, has been developed. Figure 4
illustrates the commercial device.
Equation 1.1 indicates that the settling rate decreases with increase in pulp density
(solids concentration by volume), and the term vso can be interpreted as being the
terminal velocity of particles of an equivalent Stokesian diameter; settling rate thus
increases with particle size. Both these trends of course conform with common
experience.
If it is assumed that the density below the mudline, but clear of the compaction
zone, is substantially constant, c, then we can write:
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
12
dρ v
ρc s (3.1)
dt h
Temperature affects the settling rate, due to the decrease in water viscosity
associated with a rise in temperature. The effect amounts to approximately a 10%
increase in settling rate per 4 degree Celsius rise in temperature.
Magnetisation of the media, as may occur if the demagnetising coils installed on the
medium return pipes are faulty or absent, will cause magnetic flocculation, which
can lead to higher settling rates. This is hard to quantify because the degree of
magnetisation is difficult to gauge, but as an indication, it has been found that
material which has been magnetised by stroking the test tube container with a
laboratory hand magnet has a settling rate twice that of demagnetised medium.
(Perversely, magnetic flocculation is known also to increase viscosity under some
circumstances.)
These effects make it difficult to deduce a size grading from a sample of medium
taken from a working circuit and tested by the stability meter, but, as noted earlier,
the importance of the stability figure is that it reflects the behaviour of the working
medium more fully than a dry size grading alone, taking account of effects which
would operate in the circuit.
A table showing typical settling rates found for two commercial grades of ferrosilicon
is shown below, and the results are depicted in Figure 5.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
13
The stability meter has now been in routine use on a large DMS plant for several
months. The dense medium cyclone circuits consist of 8 primary cyclone modules
using 610 mm cyclones, which treat the -25 +3 mm product from crushing and
milling. There is one 400 mm cyclone circuit, which upgrades the combined
concentrate from these circuits, and a pair of 350 mm cyclones, which upgrades the
gravity concentrate of the -3 +0,5 mm fraction drawn from a separate
preconcentration plant. The primary circuits are labelled D to L, the coarse
reconcentrate circuit is M and the fines reconcentrate circuit is N.
A sampling scheme was adopted, which drew medium from 2 circuits per day for
stability and size distribution analyses, over a period of approximately ten weeks;
the data obtained are shown in Table 2.
CIRCUIT D E F G H J K L M N
SG 2.59 2.66 2.66 2.56 2.45 2.52 2.66 2.46 2.75 2.63
Grading
57.6 60.8 65.7 59.9 64.0 61.2 65.1 65.3 57.0 63.1
(% -44 m)
Settling Rate
11.6 9.5 9.8 9.6 9.8 10.2 6.7 9.5 26.0 31.6
(g/ℓ/s)
Standard
Deviation of 2.8 2.4 1.9 5.6 1.3 3.6 1.6 3.4 13.8 9.8
Settling Rate
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
14
When examining the data from the individual circuits, correlations between settling
rate and density, and settling rate and size grading, are clearly apparent. These
data are compared with typical values for clean media in Figure 6.
Generally the values of settling rate (or stability) of the primary circuit media were in
the range expected of media conforming to 100D grading, or slightly coarser,
allowing for approximately 1% contamination by colloidal slimes.
The size distributions of the media used in the two reconcentrate circuits were
equivalent or perhaps slightly coarser, and the non-magnetics content was 0,3%,
which can only partly explain the large difference observed in the settling rates.
Possible explanations for the greater settling rates of the reconcentrate circuit
media include flocculation because of magnetisation or some corrosion-related
effect.
The latter is mentioned because the reconcentrate media tend to be used for only
approximately 40% of the time, being in closed circulation or storage for the balance
of the time. Such a flocculation effect would not be detected by normal screening
because the sample would be dried (100 ºC) and then ground lightly in a mortar and
pestle or brushed gently through a 200 m screen to break up cemented particles
before screening. Another practice is to add small rubber balls to the top deck of the
screen nest.
Ore sampling was therefore conducted in order to define the prevailing Tromp
curves for the different circuits, to determine whether the separation parameters
differed. Specifically, the D module (primary) and M module (reconcentration) were
sampled during a period when the large difference in measured stabilities persisted.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
15
Table 4 shows an analysis of the media from the two circuits. Interestingly, the size
analysis carried out by Microtrac (laser dispersion) shows the M module medium to
be coarser, while the screen analysis on the dried powder shows it finer. This is a
further indication that agglomeration may have been taking place in the M module
medium.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
16
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
17
Table 5 shows the Tromp curve parameters in the various size fractions, and as a
total stream.
SG50 Ep
SIZE RANGE
MODULE MODULE
(mm)
D M D M
-1.0 +0.6 - +3.6 - -
-2.0 +1.0 3.13 3.36 0.20 0.141
-2.8 +2.0 3.05 3.37 0.076 0.106
-5.6 +2.8 2.96 3.34 0.071 0.094
-9.5 +5.6 2.94 3.30 0.055 0.073
-13.2 +9.5 2.93 3.30 0.045 0.073
-19.0 +13.2 2.91 3.30 0.045 0.045
-19.0 2.91 3.38 0.045 0.040
TOTAL 2.96 3.32 0.072 0.090
Cut-Point
The M module cyclone has a density cut-point significantly higher than that of the
D module. Table 5 shows that this is true at virtually all densitities, the M module
cutting 0,2 to 0,3 SG units higher than the D module. This is in spite of a less dense
feed medium. The difference in cyclone size - 400 mm vs 610 mm - is thought to be
too small to explain this effect, which is ascribed to the effect of an unstable
medium, compounded by the higher density of the feed ore to the M module.
Quality of Separation
The slope and shape of the Tromp curves point to a more ideal separation in the
D module. In contrast, the M module has significant misplaced material thought to
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
18
Stability-Viscosity Relationship
In the fine 1-2 mm fraction it appears that the quality of separation in D module was
worse than that of M module, despite the overall superiority of the D module
performance. A possible reason for this apparent anomaly is the effect of high
slimes concentration, which simultaneously enhances the stability and increases
viscosity. As a result of the improved stability, the coarse particles will be separated
with little misplacement, but the finer particles will be entrained by the viscous
medium and consequently split more in proportion to the volume split of medium to
cyclone underflow and overflow[1,24].
These results confirmed earlier tracer tests (with 4 and 6 mm tracers), which had
shown that the D module was cutting at approximately 2,7 SG while the M module
was cutting at 3,2 SG. Attention was then devoted to improving the demagnetisation
of in-circuit medium and maintaining the correct medium size distribution. The
settling rate of the M module medium was subsequently found to be more
reasonable, equivalent to 18 g/ℓ/s at SG 2,6, and this was reduced only slightly by
demagnetisation, which suggests that magnetic flocculation cannot have been the
major cause of the low stability. It is now believed that the intermittent operation,
and long periods of no feed, which characterise a reconcentration circuit, gives rise
to a loss of fine ferrosilicon preferentially, and also a drop in the level of slimes. This
has a destabilising effect.
3.5 CONCLUSIONS
Media are stabilised by the presence of colloidal slimes, but this may have a
detrimental effect in the fine ore sizes because of the concomitant increase in
medium viscosity.
Estimation of stability from mudline settling rates is not always possible and is
usually tedious. An equivalent, and in practical terms a more relevant settling rate,
can be measured by following the change of density in a defined zone of a sample
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
19
tube. The use of an inductance sensor coil makes it feasible to link this process
very conveniently to an electronic processor, which will read both the start density
and the settling rate.
This Stability Meter has a standardised geometry and stirring method, and very
straightforward operation. The use of a microprocessor permits a display, which not
only returns the results but also prompts the operator, in a choice of languages, to
continue the testwork with the correct procedures.
Calibration of the stability meter, to allow for changes in the working environment
and component drift, is achieved by inserting a calibration slug on a prompt from the
meter, and there are no knobs to turn or trimmings to be made.
With the advent of this technique, the problem of assuring the quality of the medium
in a DMS process becomes greatly simplified by comparison with the slow and
sometimes error-prone methods of size distribution analysis.
4.1 DESCRIPTION
The on-line viscometer[29] was developed specifically for use with unstable
suspensions, and is shown in Figure 7. It is of the rotational type, and consists of
two main components - a measuring head with associated electronics, and an
integral sample preparation/presentation system. Medium is pumped to a steady
head box, where tramp oversize is removed, before flowing under gravity to the
rotating bobbin, which is driven by a micromotor in the measuring head (Figure 8).
Current drawn by the motor is a function of drag on the bobbin and rotation rate.
The electronics converts this current to a voltage, which is then displayed. For plant
monitoring of medium viscosity, the bobbin is rotated at a fixed rate, so that the
reading can be related to viscosity via a simple calibration chart (Figure 9).
Calibration is carried out using fluids of known viscosity, such as glycerine.
The viscometer was installed in closed circuit with the correct medium sump, as
shown in Figure 10. As with the other instruments, the viscometer gives an output
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
20
As noted earlier, it is known that medium viscosity has a direct effect on the
performance of a DMS cyclone. As part of a larger investigation, the opportunity
was therefore taken to examine the effect of viscosity on cyclone performance,
using the plant just described.
A series of tests was conducted at three different viscosity levels, using the tromp
partition curve as the performance criterion. The medium was 100D milled grade [13]
and its viscosity was adjusted by the addition of fine bentonite clay directly to the
correct medium sump. The size distributions of the bentonite and ferrosilicon were
measured using a Microtrac particle size analyser and are given in Table 6.
An example of the instrument's output, logged by the computer, shows how the
viscosity is affected by bentonite addition (Figure 12). Viscosity rises to a peak as
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
21
the clay becomes mixed, after which it decays as the medium is cleaned by the
magnetic separators. Thus, the two tests at high viscosity were run over periods of
viscosity change and not at fixed values. Non-Newtonian behaviour of the medium
cannot easily be measured on-line and was therefore not monitored during this
work. For this reason viscosity is quoted in this paper as the "equivalent apparent
Newtonian viscosity", i.e. the viscosity that a Newtonian fluid would have for a given
viscometer reading - see Figure 9. Experience has shown that the error in shear
rate due to non-Newtonian behaviour in ferrosilicon varies, but can be as high as
15%.
Test conditions, other than viscosity, were fixed as far as possible to conventional
levels. A standard 356 mm x 20º cyclone, fitted with a 102 mm spigot and a 146 mm
vortex finder, was fed at nominally 180 kPa inlet pressure with 2,65 SG medium.
The actual conditions logged by the computer during each test are given in Table 7
and the salient features of the Tromp curves are summarised in Table 8.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
22
CONCENTRATE
% OVERALL 2.4 3.0 9.1
YIELD
-2.0 +1.0 mm 3.264 3.192 3.049
-2.8 +2.0 mm 3.160 3.087 2.961
CUT-POINT SG Units -5.6 +2.8 mm 3.144 3.097 2.943
-8.0 +5.6 mm 3.192 3.112 2.925
Overall 3.176 3.128 2.961
-2.0 +1.0 mm 0.080 0.094 0.085
-2.8 +2.0 mm 0.093 0.040 0.088
EP (75-25) SG Units -5.6 +2.8 mm 0.073 0.056 0.064
-8.0 +5.6 mm 0.110 0.072 0.055
Overall 0.093 0.072 0.080
-2.0 +1.0 mm 7.402 9.119 7.705
ERROR AREA -2.8 +2.0 mm 8.648 5.072 8.196
(TOTAL) -5.6 +2.8 mm 8.005 6.760 5.302
-8.0 +5.6 mm 9.479 7.552 4.025
Overall 9.166 8.590 7.187
-2.0 +1.0 mm 1.21 1.49 4.76
MISPLACED % -2.8 +2.0 mm 1.30 1.69 3.20
LIGHTS of Feed -5.6 +2.8 mm 1.65 2.42 3.60
-8.0 +5.6 mm 1.86 2.53 3.13
Overall 1.49 2.19 3.72
-2.0 +1.0 mm 0.04 0.17 0.29
MISPLACED % -2.8 +2.0 mm 0.19 0.05 1.50
HEAVIES of Feed -5.6 +2.8 mm 0.14 0.05 1.10
-8.0 +5.6 mm 0.15 0.07 0.64
Overall 0.14 0.11 1.36
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
23
WEIGHT %
SG INTERVALS
IN SG INTERVAL
-2.5 0.41
-2.7 +2.5 51.57
-2.95 +2.7 38.10
-3.1 +2.95 8.26
-3.3 +3.1 0.95
-3.5 +3.3 0.15
-3.6 +3.5 0.06
>3.6 0.50
The first test, run without bentonite at an apparent viscosity of 23 mPa s, ( 23cP)
produced results representative of normal cyclone operation. The overall cut-point
was 3,18 SG, with a concentrate yield of 2,4%. The viscosity in Test 2 was
increased initially to 114 mPa s bentonite addition and had fallen to 63 mPa s by the
end of the test. This increase had the effect of lowering the cut-point from 3,18 to
3,13, as expected from previous experience, which increased the concentrate yield
to 3,0%. In Test 3 the viscosity was increased further to 192 mPa s at the start of
the test, falling to 108 mPa s at the end. The overall cut-point was now reduced
further to 2,96, which was located within a large quantity of near-density ore (see
the ore density distribution in Table 9). A proportion of this ore consequently
reported to concentrate increasing the yield to 9,1%. This represents a nearly
fourfold increase in yield over the base case (Test 1) and obviously spells disaster
for any downstream processes, as they would then be completely overloaded.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
24
In conclusion, the viscometer has been shown to permit the apparent viscosity of in-
circuit medium to be monitored on-line. This facility allows the plant operator to
anticipate separation problems caused by increases in viscosity. Remedial action
(cleaning more medium, improving the washing of the feed ore, or improving the de-
magnetisation of the medium) can then be taken, either manually or, where
appropriate, automatically using the 4-20 mA signal available from the viscometer.
As part of the production trials of a series of prototype viscometers, one unit was
installed on the float medium return line of a production 4,9 m diameter DMS cone,
treating 150 t/h of ore. The purpose of this trial, which lasted five months, was to
gain experience in the problems associated with sampling medium for the
viscometer, and to assess the robustness of the instrument. The lessons learned
have been incorporated into production versions of the instrument.
A sample of correct density medium from the cone circuit was not available, as the
underflow and overflow media gravitated to separate sumps and returned to the
feed independently. Consequently, and for practical reasons, the viscometer was
installed on the float medium return line. The cone operates using atomised
ferrosilicon at a feed SG of 3,15. The differential between float and feed medium
SG is small, and therefore it was considered that viscosity changes in feed medium
would be fairly reflected in the float medium. The instrument reading was recorded
continuously on a 24-hour chart recorder for later analysis.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
25
It was concluded from the testwork that careful attention must be paid to sampling
and sample preparation in production applications, and that with a satisfactory feed
the instrument provides useful information about the condition of the medium.
It has been shown that the rheology of a dense medium can be characterised by its
stability and its apparent viscosity, and that both these properties, independently
and in association, are an important influence on the performance of a dense
medium cyclone. It is therefore desirable to have methods of providing meaningful
and reliable measures of both quantities in a plant environment.
A stability meter has been developed, which utilises an inductance coil to provide
an off-line estimate of the rate of change of density in a medium, due to settling at a
point in a sample tube. Digital processing using a microprocessor permits a simple
and fast measurement procedure, free of operator bias, and the output of a stability
figure expressed in absolute units of g/ℓ/s. It is emphasised that the stability is a
more meaningful reflection of the condition of the medium than conventional
analytical procedures such as the measurement of medium size distribution. A case
study has demonstrated that the routine monitoring of stability can identify
conditions, which will result in a deterioration or change in cyclone performance not
easily detected by other means.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
26
Application of the stability meter is, by the nature of the principle of measurement,
restricted to ferromagnetic slurries such as ferrosilicon or magnetite media. The
viscometer is however appropriate to any stable or unstable fine particle
suspension, and consideration is currently being given to its application in the on-
line measurement of the rheology of grinding mill discharges and thickener
underflows.
ACKNOWLEDGEMENT
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
27
REFERENCES
3. Van der Walt, P.J. & "Determination of the Viscosity of Unstable Industrial
Fourie, A.M. Suspensions with the Aid of a Stormer Viscometer".
Min. Met., July 1957, 709-723.
7. Van der Walt, P.J. "A Method for Determining the Settling Rate of Heavy
Fourie, A.M and Medium Suspensions".
Van Doornum, G.A.W. Fuel Research Inst., Pretoria, June 1959.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
28
19. Lilge, E.O. "Apparent Viscosities of Heavy Media and the Driessen
Fregren, T.E. and Cone".
Purdy, G.R. Trans. Inst. Min. Met., 67, 6, 1957-58, 229-249.
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
29
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
30
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
31
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
32
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
33
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
34
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
35
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
36
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
37
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
38
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants
39
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology in DM Cyclone Plants
40
nd
2 Samancor Symposium on DM Separation - 1985: The Monitoring of Medium Rheology
in DM Cyclone Plants