eo
MINERAL PROCESSING
The application of residence time distributions
to flotation and mixing circuits
J.E. NESSET
Mattabi Mines Ltd.
Ignace, Ontario
ABSTRACT
This paper discusses the importance of considering residence
time distribution (RTD) in the design of flotation and mixing
circuits and in the analysis of operating problems, The experi-
fences and successes of Brunswick Mining in this area are
covered and include: the development of @ practical RTD
‘measurement technique using pulp conductivity, baffle installa
tion o improve the RTD of existing process equipment, and the
‘method in which RTD considerations have been used to design
the newest flotation and conditioning operations.
Introduction
‘The use of residence time distributions (RTDs) in the design and
analysis of unit operations is well understood by the chemical
engineering profession' while mineral processing engineers
hhave tended to rely on empirical and rule-of-thumb factors. For
example, in scale up from laboratory and pilot test results to full
size flotation plant design, empirical rules are often used to
establish residence time and number of cells in the bank. Cole-
rman® suggests a factor of at east 2, and up to $ for gold ores,
be applied to laboratory determined flotation times. This is fur-
ther illustrated in Table 1, taken from Denver equipment”).
‘The number of cels in Series to achieve the required retention
time while minimizing shor-cicuiting has generated a variety of
conclusions” (Table 2). Obviously there is less than complete
ALE. Nesset
Jan Baward Nesset recived ¢ BEng.
(metallurgical) from McGill University
stitution in I9BD. From 1979 1986 Jan
[Nesset was employed with Norands's
Brunswick Mining. operation near
Bathurst, New Brunswick, as concen-
‘rator metallurgis, project meallurgst and senior metallrgst. Since
1987 he has Bee wil superintendent at Noranda's Mattabi Mines oper
ation near Ignace, Ontaria,
‘Mr. Nesset has authored numerous papers relating to magnetic
property measurement and high gradient magnetic separation of
{aconites and sulphides, pressure filtration of concentrates, ss well a5
trinding and flotation operations,
“He ise member of CIM,
Keywords: Mineral processing, Flotation, Residence time dstibuton
‘agreement on the subject. Young notes that a clear difference
of opinion has emerged and cites as extremes, () the publica-
tions of Outokumpu® showing unimpaited recovery for its
largest machines (38 m') compared to equivalent retention times
for numerous smaller machines, and (i) the experiences of
Eastern European countries where the practice iso install banks
‘of 30 units or more, operating in series, often with unit retention
times of only a fraction of a minute.
‘The empirical factors used by engineers presumably account
for other than tested feed conditions, inefficiencies in large con-
tinuous machines, inevitable day-to-day circuit disruptions,
sradual equipment deterioration and any other safety factor the
engineer feels compelled to incorporate. An understanding of
RTD concepts and techniques should help the practicing mineral
engineer to develop more soundly based factors. This paper
relates the experiences of Brunswick Mining in using RTD tech-
niques in the unit operations of flotation and mixing.
Theoretical Aspects
Consider a process vessel of volume V, being fed a slurry at a
volumetric flowrate, Q. The slurry mean residence time in the
vessel, s gven by:
i- vg
‘The Muid and mineral particles do not all stay in the vessel the
same length of time; some may exit after a very short time
(resulting in short-cireuiting) and others remain for a much
Tonger time. This can be represented by a “residence time
distribution”* or RTD, which is simply a probability function
against time.
‘The simplest method of determining the RTD of a vessels by
the stimulus-response method in which a tracer is introduced to
the vessel infet stream at time zero. The outlet stream is
‘monitored either continuously or discretely to detect how
passage through the vessel has distributed the input signal. The
simplest response to analyze is from a pulsed input, Figure |
shows an RTD curve, resulting from a pulsed input, measured
concentration, C, "in the vessel exit stream.
‘An RTD curve is characterized by it first two moments, i the
‘mean residence time, and ot, the variance. The mean residence
time reflects mass transport phenomenon in that it defines the
relationship between the vessel volume (V) and volumetric
feedrate (Q). The variance, in turn, reflects the degree of mix-
ing which occurs within the vessel. Both may be readily deter-
‘mined from discrete measurement values by the following
Approximations. Measured concentrations are fist normalized
so that the total area under the curve equals unity, ie.
Paper reviewed and approved for publication by the Canadian Mineral Processors Division of CIM.
CCIM Bulletin, November 1988
8eel pulse npat
Tracer cutout et euve
Ke
ator ZEA
where
CC = normalized concentration
E = measured concentration
24, time interval
‘The mean residence time is found from:
where
fea ecm ge abit!
Fea * EG 3% av
= Ey Cay since EC abe
“The variance is found from:
_ fice 4 EG ay
cee ca
= ERG ay - 2
discrete time value
srete concentration value
For several closed vessels in series, the over-all mean residence
TABLE 1. Typical flotation residence times compar-
ing laboratory and industrial conditions (Ret. 3)
Usual industial Denver lab col
Material flotation times (min)
{Hlotable form) times (min)* _ (0:2 or 0-12 cel)
Bante B40 45
Coal 35. 23
Copper 13.16 oa
Effluents e12 a5
Fluorspar B10 a5
Feldspar 8:10 34
‘Lead 68 35
Molybdenum 1420 87
Nickel 1014 67
oil 48 23
Phosphate 48 23
Potash 48 23
Sand (impurity flotation) 7.8 34
Silica (ron ore) 810 35.
a (phosphate) 45 23
Tungsten B12 56
‘ine. B12 86.
“For roughing applications; fr cleaning, he rotenton time ie between
‘6087880 ofthe roughing time fora particular mineral, thee Side
15 50%-05% ofthe normal roughing soli
6
FIGURE 2. RTD cures for varios degrees of dspesion as
‘measured by the D/UL parameter (rom Levenspe)
time and variance are simply the sum of the individual values:
pred = Ee
It is sometimes convenient to express these parameters in
dimensionless time units:
G =i
a
y= “(the reative variance)
with the mean dimensionless residence time equal to unity by
Sefinition:
‘There are two types of ideal flow; plug flow in which the
‘uid stream passes through the vessel without mixing with the
vessel contents, and fully mixed flow in which the fluid ele
ments are immediately dispersed throughout the tank volume.
‘Actual flow has components of both types of ideal low.
Most modelling of RTD curves has made use of either dis
persion or tanks-in-series models. A thorough discussion of
the various models and their applicability is covered in the text
bby Levenspiel\0. The degree of mixing has been quantified by
Levenspiel for closed vessels through the application of @
dispersion model and a dimensionless group called the "vessel
dispersion number", D/UL, a parameter which measures the
‘extent of axial dispersion and where
D = axial dispersion coefficient
L = length of vessel
U = anil id velocity in vest
For elosed ves, Levens! found
og = So zorwL-210/UL PA - et-ULID)
forthe famly of curves shown in Figure 2.
'A plot of the preceeding equation appears in Figure 3 and
may be sed determine B/UL values for experimental RTD
caves one tro have bom cst” Aword of can
isin order thse relaonsip ar for closed vessel (id
Eaters and eaves solely by pug ow) and lose thee apical:
iy where very lage amounts of dispersion sre preset
Volume 81, No. 919FIGURE 3, The relative variance asa function of D/UL after
Levenspid,
Nevertheless, itis possible to use the D/UL parameter to
predict the extent of reaction (or recovery) which will occur.
Figure 4 has been developed by Levenspiel, assuming first-
order reaction kinetics, and relates the real to plug flow (or
batch) volume, V/V,, D/UL and the extent of the reaction,
Xq (expressed in the figure as I-X,),
An Example: Lime Slaking
I it has been determined in the laboratory that batch lime
slaking requires 15 minutes forthe reaction to go to 98% com-
pletion, what mean residence time should be designed for two
Inseries mixers in a continuous plant operation so as to
achieve the same extent of reaction? Assuming that through
cither prior RTD measurement on single mixers or reference
‘material one knows that o,, the relative variance, is 0.8, use is
‘made of the additive nature of the variances to calculate the
expected og? for two units, namely 0.4. Use is then made of
Figure 3 to obtain the corresponding D/UL (=0.3). For an Xq
of 0.95 and a D/UL of 0.3, Figure 4 shows that a real to plug
flow volume (or time) ratio of approximately 1.8 will be re-
4uired. The two plant mixers should, therefore, be designed to
provide a total mean residence time of 27 minutes (1.8 x 15
minutes).
Dobby and Finch” have applied the preceeding method of
reactor scale-up, assuming frst order kinetics, to column flo-
tation with considerable success. The reader is again referred
to the text by Levenspiel which covers the concepts and appli-
FIGURE 4. Comparison of real vs eal flow (ue. plug flow)
equieed volumes, V/Vp, (rom LevenspieD.
cations of RTDs in detail and with great clarity. His chap-
ter on diagnosing ills of operating equipment should
prove particularly enlightening to plant operators. In sum-
‘mary, it need only be said that RTD concepts are easy to grasp
and apply and that the results can be interpreted both qualita-
tively as well as quantitatively. At Brunswick Mining much of
this work has been carried out by engineering students
Measurement of Residence Time
Distribution
Measurement Techniques
RTD measurement involves the introduction of a tracer to the
vessel feed stream and either periodic sampling (discrete
‘method) or continuous monitoring (continuous method) of the
tracer in the vessel discharge. In a processing plant one is most
often concerned with the behaviour of the solids. The use of
solid tracers, though by no means impossible, does tend to add
to both the complexity and cost of the test. The practice at
Brunswick Mining has been 10 use liquid phase tracers and to
‘make the assumption thatthe liquid phase ofthe slurry closely
approximates the behaviour of the solids phase. Dobbyt!® has
shown thatthe validity of this approximation is a function of
Particle size and improves with decreasing particle diameter.
At the very fine sizes present in the Brunswick flotation circuit,
TABLE 2. Literature sampling on the required number of flotation cells in series
Recommended
number of
Publication calls ‘Additional
Roterence et In sories Comments
Sutherland, O.N. ‘A study of the optimization of the arrange 68 Usually sufficient
ment of flotation circuits (4)
Churchill, M Large flotation machine development and 78 For typical sulphides
Degner, WA. ‘operation (8)
Abiter, Ny Design of flotation cells an circuits (6) 1244 12 minimum
Weiss, N,
Wilson, R.A. 1218 12 minimum
Dawson, HA
Nit Experiences with large Outokumpu flotation 6s 4 minimum
Tarvainon, M machines (8) for large cells
itu, Discussion period 45
Degner, V.A, Authors’ replies, 63
Dominici, P XiVIM PC. 1012 Minimum
Lawrence, GA. Toronto, Canada (1982) (9) 48
CCIM Bulletin, November 1988
”