FURTHER DEVELOPMENTS IN THE CONTROL

OF LEVEL IN FLOTATION COLUMNS

M H MOYS

DEPARTMENT OF CHEMICAL ENGINEERING

UNIVERSITY OF THE WITWATERSRAND
 Further Developments in the Control of Level

in Flotation Columns

M. H. Moys,
Consultant in Column Flotation,
Multotec Cyclones (Pty) Ltd.
and
Department of Chemical Engineering
University of the Witwatersrand
Johannesburg

Abstract
Interface level in flotation columns is generally measured with the
aid of a pressure-sensitive instrument ego a differential pressure
cell or Metritape. The use of this type of measurement is critisised,
and other methods which are claimed to give more accurate measures of
interface level are discussed. The paper focusses on the development
of methods based on the measurement of conductivity across the froth-
-slurry interface. These are generally sabotaged by variations in the
conductivity of the feed slurry, particularly when the pH of the
slurry is being controlled, ego in sulphide flotation. Methods of
improving the accuracy and reliability of this technique are devel-
oped. The result is a level control scheme which successfully rejects
large disturbances in plant operation (including a 5:1 change in
slurry conductivity) while controlling level to an accuracy of ± 4cm.

1. Introduction

For effective operation of a flotation column the level must be

controlled near a pre-specified level L set ' so that the froth height
hf is (a) not so low that the significant mixing that occurs in the
froth phase results in a loss in grade, Hnd (b) not too high, n~stll­
ting both in an excessive residence time of bubbles in the froth phase
and a loss of residence time in the pulp phase (both of whi~h effects
result in a loss of recovery). Yianatos et. al. (1987) recommend a
froth height of approximately 1 m. Current methods for measuring
interface level L are reviewed below, and research work aimed at
developing a reliable method of measuring L using conductivity
measurements is described.
 Level Control in Column Flotation Moys Page 2

2. Review of methods for measuring interface level

The method most widely used is that based on a measurement of hydro-

static pressure P at a distance between 1,5 and 2 m below the overflow
lip. P is however not an accurate indicator of interface level because
it is a strong function of other variables, such as slurry density,
hold-up of liquid in the froth, bubble loading, etc. Moys and Finch
(l988a) showed that errors in L exceeding 0,5 m are easily possible
when high gas rates, small bubbles and high bubble loadings occur,
particularly when the flotable mineral is dense, e.g. pyrite or
galena. Moys and Finch (1989) measured an error of 0,4 m in galena
flotation in a limited test programme at the Polaris Mine, N.W.T.,
Canada and the variation of level with pneumatic pressure measurement
in pyrite testwork in South Africa is shown in Figure 1. This lack of
precision means that frequent checks on level must be made by the
operator in situations where a sight window on the side of the column
exists; in the absence of a sight window the poor performance of the
measurement cannot be checked, so the operator is forced to set the
froth depth at a conservatively high level to avoid disaster, thereby
incurring losses associated with unnecessarily deep froths most of the
time.

Several papers have reported the use of Metritape, a patented device

which consists of a wire coil wound round a contact strip and insula-
ted from the slurry by a flexible, corrosion-resistant sheath. A strip
of Metritape 2m long is inserted in the top of the column. The design
is such that the wire does not make contact with the contact strip
until the pressure in the fluid outside the tape exceeds l50mm water.
This is sufficient to collapse the sheath and wire coil so that
contact is made. The resistance of the wire which is not in contact
with the contact strip is measured and provides and accurate (± 1 cm)
measure of the level at which the first contact is made. ~~ile Metri-
tape is widely used for the measurement of fluid level in tanks, its
use in colu~n flotation must be seriously questioned, since it
provides a measure of the depth in the column at which the pressure is
150 mm of water above atmospheric pressure. This can only provide
relatively accurate measurements of interface level for columns where
the pressure at the interface level is less than this critical value -
a condition which can only occur for froths which are lightly laden,
have little interstitial fluid and/or where froth height is low. The
measurement is in addition subject to all the disadvantages of dip
cells discussed above.

Accurate measurement of interface level relies on the measurement of

intensive properties of the fluids near the interface (such as densi-
ty, temperature or conductivity) rather than an extensive property
such as pneumatic pressure which is a function of the total mass of
 Level Control in Column Flotation Moys Page 3

material obove the measurement point. Some of these are discussed

below.

Techniques based on the measurement of the position of a float at the

interface (this is essentially a density technique and is used
successfully with conventional flotation) cannot be relied upon since
the density of the froth near the froth-slurry interface varies over
too wide a range. It is not generally possible to select the density
of the float which will apply to all possible conditions.

Moys and Finch (1988a, 1988b and 1989) have reported the use of
measurements of temperature profile across the froth-slurry interface.
The technique relies on the existence of a significant and reliable
temperature differential between the wash water and the feed slurry.
Provided that the column is operated under positive bias conditions
(i.e. with wash water flowing downwards through the froth phase, so
that the froth phase temperature is close to the wash water tempera-
ture) there is a sharp change in temperature at the interface.
Measurements obtained in an industrial galena cleaner are shown in
Figure 2 (Moys and Finch, 1989). Detection of the level at which this
change occurs provides an estimate of interface level accurate to
± 7 cm. The method suffers the disadvantage of requiring intelligent
analysis of 10-15 temperature measurements, so will be expensive.
Significant other advantages (see Moys and Finch, 1989) to be obtained
from such measurements (such as online measurements of froth washing
efficiency) may make this technique an attractive option, particularly
in an environment where computer control resources are already avail-
able.

The use of conductivity measurements may provide a measurement which

is reliable, simple to implement and cheap. Moys and Finch (1988a)
discussed initial work in this area, using the conductivity sensor
arrangements shown in Figures 3 and 4. The first involves the use of
two paralllel, vertical probes which provide a fairly linear correla-
tion between level and conductivity. The second involves measurement
of conductivity between two probes, one above and one below the inter-
face. The correlation is non-linear but the simplicity of the design
(the two probes can be mounted on a vertical rod passing through the
froth phase) makes the exploitation of this approach attractive.

Multotec Cyclones (Pty) Ltd (a South African supplier of flotation

columns) have used this method successfully on pilot column tests at a
local fluorspar mine, and it has also been used in coal column flota-
tion tests (sponsored by the National Energy Council) at the Universi-
ty of the Witwatersrand. The conductivity signal is connected to a PID
controller with a conductivity setpoint which keeps the interface just
below the lower edge of the upper conductivity probe. Level was
 Level Control in Column Flotation Moys Page 4

controlled to an accuracy of ± 1 cm.

While the method worked well in the above applications it failed when
applied to pyrite flotation, because the conductivity of the slurry
varied widely as a result of the necessity to control the pH of the
slurry at a setpoint of 3,8 pH units. Small variations in pH resulted
in large variations in conductivity, so it was impossible to select a
reliable setpoint for the conductivity (i.e. level) controller. It was
necessary in this situation to revert to the use of pneumatic measure-
ments for level control. The data given in Figure 1 was collected
during this experimental programme: clearly a more reliable estimate
of interface level is required.

We decided to invetigate further the use of conductivity measurements.

It was evident that a control scheme was required which would adapt to
the variations in slurry conductivity. The development of such a
scheme is described below.

3. Theoretical Developments

3.1 The variation of conductivity across the interface.

This is discussed by Moys and Finch (1988a) in some detail; the

salient points are summarised here.

The conductivity of a gas-liquid dispersion typical of column flota-

tion froths (Yianatos, et al, 1985) is given by

- 0,432 kl,fr

\vhere kl , fr and €l,fr are the conductivity and fractional holdup

respectively of the liquid between the bubbles in the froth phase.
Typically ~l,fr is leGs than 0,4 at the base of the froth and drops
rapidly to approximately 0,2 at
the top of the froth phase. kfr , 1
will be similar to the conductivity of the slurry ksl at the base of
the froth but in columns where washing of the froth phase by the wash
water is efficient it will change rapidly towards the conductivity of
the wash water.

In the pulp phase the conductivity is given by

k
pulp 1,55 - 0,55€sl

where slurry hold-up €sl is generally greater than 0,7. For kl , fr

 Level Control in Column Flotation Moys Page 5

k sl ' €l,fr 0,4 and €sl 0,7 the ratio between froth and
pulp conductivities is 0,48. In cases where ~ash < ksl this ratio
will be decreased, giving ample variation in k across the interface,
while if ~ash > ksl (e.g. this can occur in cases where the pH of
wash water must be controlled) the variation of k across the interface
will not provide a reliable basis for interface detection.

3.2 A model for conductivity across the interface

Conduction in fluids between two small probes a significant distance

apart is a complex phenomenon. In an infinite uniform medium the
conduction paths follow lines through the medium that constitute the
lines of steepest descent in the voltage field set up between the two
probes as shown in Figure 5(a) (please note that the diagrams in
Figure 5 are not accurate; they were created for illustrative purposes
only). While most of the current follows a relatively short, direct
path between the probes, a significant proportion of it follows much
longer paths. The conductivity between the two probes is a stong
function of the specific conductivity of the liquid, the distance
between the probes and the surface area and the geometrical arrange-
ment of the probes. The very simple model that assumes that the
current is conducted through two resistances set up by the froth and
the pulp provides the following model (formulated in terms of resis-
tances; see Figure 5(b»:

R
tot +
where R is resistance per cm, L is the interface level and Ztop and
Zbot are the probe positions. This implies a linear relationship
between Rtot (=l/k tot ) and L. This is true only for L some
distance from the top probe, as shown in Figure 4. A more comprehen-
sive model must account for the geometrical design of the sensors and
must in particular account for the different conduction paths followed
by the current when the interface level is rising past the probe
faces. This is illustrated in Figures 5(c) and (d), where the presence
of an interface at the top probe introduces a distortion in the paths
followed by the current between the probes.

3.3 The measurement of conductivity

Relatively cheap conductivity meters rely on the (simplified) circuit

diagram shown in Figure 6. A high-frequency alternating voltage E is
applied to a range resistor RR connected in series with the conducti-
vity sensor (which
has a resistance Rk ). The voltage drop VR across
RR is easily related to Rk :
 Level Control in Column Flotation Moys Page 6

In th~ conductivity meters we used, this voltage was converted into a

4-20 mA signal ik
corresponding to a two-decade range on k (=l/R k )
and a 0-100% guage output p as follows:

i k (mA) 4 12 20
k (0- 1 ) RR/ 10 RR 10RR
p(l) 0 50 100

Thus the meter output is strongly and non-linearly related to k;

the relationship is such that a 10% change in k results in a large
change in p or i k , relatively independent of the value of k. This
means that the signal is sensitive to changes in k over a twodecade
range of k, something which could not be achieved with a linear
instrument. The design of this instrument lends itself ideally to the
purpose of this investigation as will be revealed below at the appro-
priate dramatic moment!

4. Equipment

Experiments were performed in a square (12 cm x 12 cm) perspex labora-

tory column 1,8 m tall. The pilot plant is illustrated in Figure 7.
Two sources of feed material were available (water only; one source
was dosed with HCl to raise its conductivity substantially above that
in the other source, which was ordinary tap water). Turbine meters
provided measurements of these flowrates. Wash water dosed with
Dowfroth 250 was added at the top of the column. Gas was added through
a canvas sparger; both wash water and gas rate were measured using
rotameters. Two LTH DCM-l conductivity meters and an analog PID
controller were available. All measurements were interfaced to an
IBM-compatible PC which was programmed to display, print and store all
data collected.

5. Experimental Results and Discussion

The objective of the work was to develop a level control method which
would adapt to changes in feed conductivity. Several approaches were
investigated.

5.1 Setpoint obtained from measurement of pulp conductivity.

2 Conductivity sensors were used. Each sensor consisted of 2 brass

strips 1 cm long clamped around a 6 mm diameter perspex rod 30 cm
 Level Control in Column Flotation Moys Page 7

apart. The second conductivity meter was connected to a sensor approx-

imately a,6m below the level sensor. This was used to obtain a
measurement of pulp conductivity which was then passed through a vari-
able resistor to obtain a remote setpoint for the level controller. It
/

was found that there was excessive interference between the two
conductivity meters; this was reduced by making sure that the top
probe of the bottom sensor and the bottom probe of the top sensor were
connected to the earth connection of their respective meters, and
finally overcome by increasing the size of these probes with a wire
grid that spanned the crossectional area of the col~~n. This ensured
that no current could bypass these probes and travel between the two
non-earth probes of the sensors. Secondly it was found that matching
the sensors (both in terms of the large scale geometry of the sensor
design and in terms of the size of the probes used) was critical to
the performance of the control system. This arrangement was used
successfully to control level for a wide range of operating variables
(feedrate, gas rate and wash water rate) but failed when large changes
(e.g. 3:1) in feed conductivity were imposed.

5.2 Use of a modified single conductivity meter

The problems found in 5.1 above were probably caused by the need to
use two conductivity meters plus associated sensors which interfered
with each other and were difficult to match accurately for the whole
range of conductivities used in the these tests. An examination of the
circuit diagram for the meters led to a dramatically simplified solu-
tion to the problem. This consisted of using the resistance across the
second probe discussed above (used for measuring the pulp conductivi-
ty) as the range resistor for the first conductivity meter, as shown
in Figure 8. The governing equation for this meter then becomes

where 'Y

This arrangement eliminates problems associated with both unmatched

sensors and unmatched meters and obviously simplifies the whole
control system.

One of the existing sensors was modified to provide the sensor shown
in Figure 9. This contained only three probe surfaces, with the
centre probe being used to provide the VR signal rather than the more
general probe illustrated in Figure 8. Existing terminal connections
on the conductivity meter.were used. The measurement was now a 4-20 mA
signal denoted iq (to distinguish it from i k ) which was recorded as
a variable q (0-100% of span). This signal was connected to the
controller which was given a manual setpoint qsP = 15-20% as shown
in Figures 10 and 11 below.
 Level Control in Column Flotation Moys Page 8

To characterise the measurement the signal q was recorded as the

interface was allowed to drop over the full height spanned by the
level sensor. The variation of interface level L with q (and 1) is
shown in Figure 9. Clearly the sensor design was not symmetrical with
respect to the middle probe surface; if this had been the case, then
we would have obtained q - 50% when the sensor was completely submer-
ged (L> 50 cm.), because in this case we would have had equal conduc-
tivities (1 - 1) between the two probe pairs. No adjustments were made
so that the necessity for probe matching could be assessed.

The ability of this arrangement to provide adequate control of level

in the face of operating changes is shown in Figure 10. Clearly
changes in
the gas superficial velocity J G provide much more serious
disturbances than changes in the wash water J W. Nevertheless the
control scheme controlled level near 38 cm with maximum deviations of
4 cm.

The effect of a setpoint change and a major disturbance in feed

conductivity is shown in Figure 11. The setpoint response overshoots
slightly but is nevertheless completely satisfactory. The conductivity
disturbance produces very little effect on the level, in spite of the
fact that conductivity changed by a factor of approximately 5:l!
Extensive tests of various combinations of these disturbances showed
that the control scheme was able to control level successfully in the
face of disturbances which are generally much larger than those to be
encountered during normal operation of flotation columns.

The main problem that needs a solution before a rugged industrial

instrument becomes available is the selection of materials of
construction for the probe which will resist the corrosive environment
found in different applications while maintaining a repeatable
response over long periods of time. Reliability is more important than
accuracy where level control is concerned, and the differential
pressure cell is a tough competitor.

6. Summary and Conlusions

Various methods for level measurement have been reviewed. The work-
horse of industry, the differential pressure cell is reliable but
inaccurate. An instrument based on the measurement of conductivity has
been developed which can provide accurate control of level under a
wide range of operating conditions and changes in feed conductivity.
The method must now be tested in an industrial environment to provide
an instrument which will eliminate the need for the frequent operator
attention which is typical of existing techniques.
 Level Control in Column Flotation Moys Page 9

7, Acknowledgements

This work has been funded by a grant from the Chamber of Mines Block
Grant made available in 1988, and by Mu1totec Cyclones (Pty) Ltd. The
assistance provided by Mr G Gloag in setting up the experimental
equipment and running the experiments is gratefully acknowledged.

8. References

Moys, M.H. and Finch, J.A., 1988a. The Measurement and Control of
Level in Flotation Columns. Presented at the Int. Symp. on COlU~l
Flotation, SME Annual Meeting, Phoenix, Arizona, January 1988.

Moys, M.H. and Finch, J.A., 1988b. Developments in the Control of

Flotation Columns. Int. J. Hiner. Process., 23 : 265-278.

Moys, M.H and Finch, J.A., 1989. The use of Temperature Measurement in
the Analysis and Control of Flotation Columns. Presented at the "Role
of the Practical Metallurgist" Symposium, Mine Metallurgical Managers
. Association of South Africa, Johannesburg, June 1989.

Yianatos, J.B., Finch, J.A. and Laplante, A.R., 1987. The Cleaning
Action in Column Flotation Froths, Trans. I.H.H., 96: C199-C205

Yianatos, J.B., Laplante, A.R. and Finch, J.A., 1985. Estimation of

Local Hold-up in the Bubbling and Froth Zones of a Gas-liquid Column,
Chem. Eng. Sci., vol 40, no la, pp 1965-1968.
 MULTOTEC CYCLONES A
COLUMN FLOTATIOI~ V

BO
/1i!I III
/'
III III
III III
/'
/'
/'
90 /
III
/' fill
III
fill I!I
III
III 1/1
III
III III
100 /1Il III
/' III III
/' /Ill III
/' ill
/' III III
III
E III III
~11 III 111111
:r: III E
u
f- III III III
n. III
W
0
III ...
N

~
0
0::
"-

/'
/'
/'
/'
/'
/'
III /'
/'
/'

L-___________________________ ~ __ ~ __

460 470 4BO A90 500 510

PRESSURE READING (mV)

Figure 1. Variation of true interface level with pneumatic pressure

signal.

Frolh Surface Frolh Surface

145 \-----~---- \,\\\---\~------

\ \\ \ \ \ \
\ \\ \ \ \ \
\ \ \ \ \ \
\ \ \ \ \ \ \
\ \ \
\ \ \ \ ~ '{io +
\ \ \ , \ 20 \
\
\
\\"10\\
\
'\ ~
\
\ \
"" .
+* ...
\ 20
'\ " ,1 "Z~'" .,. \\ \\
\
\ E: "" ,,\\
E i \ " ">\ \
-" ~\~\
~ -.., '-
~ \ ~ "'..,,~
"- "- ...
-'
v> '2 '-.
'- ,,~
--"'~'. +
. E:
-------------
'-
;:t. ...
--' resl 110. 169 ...
h= 14 ...
Tesl No. 36
Jw=0.12 ... F
=
J G 1.6 J, =0.64
J w =0.12 J o =0.01
J F =0.64 ...
J o = 0.08 ~
T "mp. Probp
Locallons

(I) (11)
O.L-_______ .______________ ~L__

0.0 1.0 0.0 1.0

Normalised Temoeraturo t{J

Figure 2. Measurement of interface level by analysis of temperature

measurements obtained from probes located at the positions indicated
in the Figure.
 VCCooductlvily Signal) (Volls-') Symbol JG .lw Condlllon ~ MULTOTEC CYCLONES ~
1.4 0.2 Interlace lolling __ , COLUMN FLOTATION
o 5 10 15 20
l00~ --0--
1.4 0.2 Interlace rtslng
,
--+-- -.:::-:--..
.-..-. 1.0 0.2 ~_~~ AXIS CF
/ ~/ It S'().U.IE [F()'
1.0. 0.4 during level conlrol ,. I I ',o
,~
®
---<l)--- 1.8 0.4
Modsl Plofile /
/
/
I
I
I

I I I
®.. 951- I I I I
I I I
''0.. I I I LI --.../
15cmx1.5cm I
'11\
/.----"') Conducllvity •
I I

I• I•
_ _ _ _ _ Probes. •
I \
\
Signals \
\
JG = 1,0 crn/s 90 \ ""....
E Jw =0.2 cn/s / '" ,\'
~ / " '.....
"1'
\ ,\'.,'.,1,
-J /
<ii
~
..J
/ --. .-," j ZbOl t
,- -- ./
> 80 I
(b)
~ 85 \..l)
Probes 1 .------~-
<ll ~ /
U III
u
a
"t::
0 /
~
.s
\ "£;III .
/
/
/
/ j{'/ /'" "..-'- =: : : :. -~_ ~. -U!~ Jr
frc:OTII
- "- , >.-
--:::::::-::'>' i ~~~~~iRl'
80 . / / ,'v'" , /,,1- __ :-:::'::: 11. c"uudRr .--- ..... I
/ i' ,',8 /
/ / ,,~. 1/ -:.---:::
i , 0 ~I
Irf I
/

I
I

/
/' i

1'"
___
.\!
j I
~
~ -H-..!
!
_."........,._~.....::-:..:-:..:_~_1iJl ~ -
,. I .J
"I(RF;,o(

J .'
I I , (
" ,I
~., /: /
·,1 Vc ~ I I
I
, I
I
\+ 75
"
,
,.~ ! .~.~
~"
\
\
\ \

;11
:
'. , , ,.§, \ \
I'

<~..:~~\ ---- __ -=--:::::..

. -'=:~~ ~==-~jl 1
.', : _________ I
,3 -------:=/
"
, /'I2 cm x 25 cm <,,~~~~~~=ti:-l ~ PULP _ _ ---~//
70 1- ?/' I cj Conductivity I ;I (d)
(~ J],-P'ObSS'-- {c)

,!~- Signals
60 • 0 &-----l
,
00 0.1 0.2 o 2 .6 .8 1.0
Conductlvily Signal Vc (volts) Conducllvity Signal Vc (voltS)

Figure 3. Response of CYO probes measuring conduc- Figure 4. Response of long parallel vertical probes Figure 5. Conduccion between t~o probes: (a)bet~een tvo points in an
tivity along a vertical axis as functions of change to changes in level under different conditions; infinite medium: Cb) a simple model for two-phase conduction; (c)
in interface level, L. response of the short pair of probes at L - 70cm.- conduction between two finite probes, and (d) effect of the presence
is also shown. of two phases.
 MULTOTEC CYCLONES.
COLUMN FLOTATION
r Terminal Connections

Recorder t - - - - - - - { ) - - - - - - - - ,
or
Co m ut 0 r 1----------0---,

Cond
Probe

Vollage lo
Curr 0fll
Conversion

O--~--l

E ('lolls I\C)

FIgure 6. Simplified circuit for a conductivity meter.

LEV EL
PROBE

Figure 7. Pilot flotation test ri' g.

 MULTOTEC CYCLONES.
COLUMN FLOTATION
r~ecorder
Level ~ or
Probe ~ Cornputor

Insulators --~

Conductors --~ v voltage to

Current
Conversion

--1::r-----'--fOse i" a to rI
E (Volts AC)

Figure 8. Conductivity cicuit adapted for the measurement of

conductivity across the interface relative to slurry conductivity.

60

w
co
0
50 0:
E D...
~ -1
-1 W
W
> >
w
W -1
-1
40
w
0 ~
Lt :>
0:
f=
0
W - - - =>
l- Q
~ 30 z
0
0

20

I .
10 15 20 25
METER OUTPUT (('0)
0,2 0,3 0.4 0,5
CONDUCTIVITY RATIO, ll=kLIkSf

Figure 9. Variation of relative conductivity with interface level.

 TT
J o : 1-43 cmls J w : 0,1 -,0,3 cmls
I I J 0'. 3 ~ 1 clllls
40 I I~
r----L--~-~~ ~

rr
u 30
c I
'"
E I I I

~'-J~~~
~
....J 20

10
700 800 I T 900
I TIME (5) I
100

::0-
~ Coni roller
rr
(3 50 Output
...:
C
0
0
0 --'------"-- - _____ -LJ, ____ L __ L ________~
800 900 1000
TIME (s)

Figure 10. Level control subject to disturbances 1.n gas rate and WilSI!

water rate.

,-----------------------~---~-~------

.-.-.

40-

kSL -1>0-
/ "
;;g 30 I 30
~
I"
"
".
to

E
rr
20 ;;>0
E
~
....J I~
/
I 15
10 /
.- /

/
/
/

z 0
100 200 300
---.-
400
10
~ TtME (s)
:J ....--- - - - - - - - - - - - - -
W ["2 (high k)
t1: 5 Fl (Iow k)
a:
$:
0
....J
LL
I --- ---
I
0 I I -,
0 100 200 TIME (s) 300 400

Figure 11. ,Level control subj ect to a large vClriation in feed

conductivity.