Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

SIMULATION STUDIES ON ROBUSTNESS OF MULTI-STAGE GRAVITY CIRCUIT PERFORMANCE…

Proceedings of the XI International Seminar on


Mineral Processing Technology (MPT-2010)
Editors: R. Singh, A. Das, P.K. Banerjee, K.K. Bhattacharyya and N.G. Goswami
© NML Jamshedpur, pp. 357–363

SIMULATION STUDIES ON ROBUSTNESS OF MULTI-STAGE


GRAVITY CIRCUIT PERFORMANCE BY RANDOM PERTURBATIONS

B. Venkoba Rao and S.J. Gopalakrishna1


Engineering and Industrial Services R&D, Tata Consultancy Services Limited, Quadra II,
Hadapsar, Pune - 411 028, India
1
Department of Mineral Processing, Post Graduate Center, Sandur, Karnataka - 583119, India

ABSTRACT
Recently methods for assessing the performance of multi-stage separation by size-density bivariate partition
surface have been developed and the algorithms have been validated using air table separation data. In this
paper, a study has been carried out on the sensitivity of two stage circuit behavior to random perturbations
that may arise in a plant operation owing to changes in operating and design variables pertaining to the
separator. Among the four possible configurations of two stage separation, it has been shown by simulation
studies that the circuits wherein primary sink stream is rewashed within a secondary separator shows better
stability in terms of overall separation efficiencies for the associated perturbations. However, the circuits
wherein primary float stream is rewashed in another separator tend to fluctuate the efficiencies of the circuit a
lot for the associated perturbations. The methodology presented in this paper can be extended to any complex
circuit. Such sensitivity studies should guide in the proper selection a new circuit among various possible
scenarios, so that one would select a more robust design.
Keywords: Gravity concentration, Simulation, Random perturbations, Two-stage separation, Partition-surface.

INTRODUCTION AND MATHEMATICAL TREATMENT


Traditionally, classifier performance efficiency is studied based on particle size whereas gravity
separator performance efficiency is studied based on particle density. A unified mathematical approach
to address all separators in terms of particle size and density together has been attempted lately.
A general partition surface represents the probability of particles belonging to certain density and
size class of feed reporting to sink stream of a gravity concentrator. A typical partition surface is
shown in Fig. 1. This surface shows both positive and reverse classification of particles based on
the pivot density of the surface. For example coarse light (density) particles of mica, would report
to float stream (overflow) rather than to sink stream (or underflow). As particle size tends towards
zero size the partition coefficient approaches to a constant value irrespective of particle density.
This constant partition coefficient can be thought of as by-pass fraction of a hydrocyclone
efficiency curve. Mathematically, a partition surface can be represented by the following Eq. (1).[1]

( ( ) )
E (d , ρ) = 1 ⎡1 + erf Ad c ρ − ρ p − B ⎤
2⎣ ⎦
… (1)

357
B. VENKOBA RAO and S.J. GOPALAKRISHNA

Eq. (1) has four parameters namely A, B, c and ρp (vis-à-vis three parameters of a hydrocyclone
efficiency curve representing size separation), which need to be estimated from the sampled data
of the separator. A procedure for this has been explained in King[2] and Venkoba Rao.[3] This
partition surface is a generic one and is applicable to all particle separators irrespective of their
design profiles. The metrics of separation-indices such as cut-size and ecart-probable (Ep) can be
derived from Eq. (1).[1] The partition surface can become steep or flat based on operating
conditions. A steep surface would give a sharp separation.

Fig. 1: Illustrative plot of partition surface.

More recently, a method to evaluate the bivariate product distributions and their marginal distributions
in terms of size or density has been derived by combining feed bivariate distribution with partition
surface.[4, 5] This method is quite useful to estimate the nature of product distributions. The estimations
can be combined for several separators in a circuit to understand circuit behavior where the product
of one unit becomes feed to the subsequent unit. The marginal distributions of the product streams
can be calculated (in discretized form) as:[4, 5]
m
piC (d ) = ∑ pijC (d , ρ)
j =1
m
p (d ) = ∑ pijS (d , ρ)
i
S

j =1 … (2)
n
pCj (ρ) = ∑ pijC (d , ρ)
i =1
n
p (ρ) = ∑ pijS (d , ρ)
S
j
i =1

Where,
Eij (d , ρ) pijF (d , ρ)
pijS (d , ρ) = n m
… (3)
∑∑ Eij (d , ρ) pijF (d , ρ)
i =1 j =1

and

pijC ( d , ρ) =
(1 − E (d , ρ) ) p (d , ρ)
ij
F
ij
… (4)
∑∑ (1 − E (d , ρ) ) p (d , ρ)
n m
F
ij ij
i =1 j =1

358
SIMULATION STUDIES ON ROBUSTNESS OF MULTI-STAGE GRAVITY CIRCUIT PERFORMANCE…

TWO STAGE SEPARATION


A number of studies in literature have emphasized the need to consider staged separation circuits
for sharp product distribution.[6–8] The staged separation can provide an inherent improvement in
the sharpness of separation that is best understood by examining individual and global partition
functions. Staged separation also provides enough flexibility to operating regime of the plant and
provides tolerance to the circuit with regard to changes in feed characteristics of the ore, changes in
equipment design settings or operational variables. In almost all unit operations of mineral processing,
namely size reduction, classification, gravity concentration, flotation as well as magnetic separation
staged operations are followed with the re-circulation of middlings so as to minimize the loss of
valuables in the unrecoverable tail fractions.
Circuit configuration and optimization of the performance of an existing circuit to yield targeted
product distribution have received considerable attention. With regard to gravity concentration, the
target optimization may involve maximization of particles of specified size and/or density along
with the minimization of contaminants in the product stream so as to maximize the revenue of an
operating plant or to facilitate the downstream operations. A simultaneous consideration of
maximization of both size and density particle attributes in gravity concentrator product streams
can be achieved through bivariate representations.
The plant operating variables, equipment design variables, ore characteristics and the separator’s
response time play a vital role in achieving the targets. The operating and design variables mainly affect
the shape of the partition surface through which the effects get propagated to the product distributions.
In the absence of any established relationships between operating variables on partition surface
parameters, one can resort to random perturbations of the partition surface parameters in some specified
range to simulate the effects of random changes in design and operating variables on two-stage
separator circuit for steady state response. The same concepts can be even extended to classifier
networks such as hydrocyclone clusters of grinding circuit wherein liberation of minerals of specified
density takes place.
Fig. 2 shows the four possible two-stage two-product circuit configurations.[6, 7] It is clear that in circuits
1 and 3 the sink of separator 1 is rewashed while in circuits 2 and 4 the float stream is rewashed.

Fig. 2: Four possible two stage circuits.

359
B. VENKOBA RAO and S.J. GOPALAKRISHNA

Let the size-density particle separation efficiencies of the first and second stage separators be
represented by Eij(1) ( d , ρ) and Eij(2) ( d , ρ) as given by the following equations.

( ( ( ) ))
Eij(1) (d , ρ) = 0.5 1 − erf A1dic1 ρ j − ρ p1 − B1 … (5)

Eij(2) (d , ρ) = 0.5 (1 − erf ( A d ( ρ − ρ ) − B ) )


2 i
c2
j p2 2
… (6)

Table 1 provides the steady state expressions for calculating the overall size-density efficiency,
EijO ( d ,ρ ) as well as overall by-pass fraction, YpO , for the four circuits under steady-state mass
balance. Thus the overall efficiency depends on the arrangement of the individual separators. As
the by-pass fraction representation is independent of particle density, it is the parameter common
to both partition surface and size efficiency curve. As the by-pass fraction of the individual
separator is always a fraction, it can be shown from the relations in Table 1 that the overall by-
pass fraction of the circuit always follow the following order.
YpO ;circuit _1 < YpO ;circuit _ 3 < YpO ;circuit _ 4 < YpO ;circuit _ 2 … (7)
Table 1: Estimation of overall partition coefficients and overall by-pass fraction for the two stage
circuit configurations from the knowledge of individual separator partition surface parameters
Overall partition coefficient, Overall by-pass
Circuit configuration EijO ( d,ρ ) fraction, YpO
Circuit 1 Eij(1) ( d , ρ ) Eij(2) ( d , ρ ) Yp(1)Yp(2)

Circuit 2 Eij(1) ( d , ρ ) + Eij( 2) ( d , ρ ) − Eij(1) ( d , ρ ) Eij(2) ( d , ρ ) Yp(1) + Yp(2) − Yp(1)Yp(2)

Circuit 3 Eij(1) ( d , ρ ) Eij(2) ( d , ρ ) Yp(1)Yp(2)

(1 − E ( d , ρ ) + Eij(1) ( d , ρ ) Eij(2) ( d , ρ ) )
(1)
ij
(1− Y (1)
p
+ Yp(1)Yp(2) )
Circuit 4 Eij(1) ( d , ρ ) Yp(1)

(1 − Eij(2) ( d , ρ) + Eij(1) ( d , ρ) Eij(2) ( d , ρ) ) (1− Y (2)


p
+ Yp(1)Yp(2) )

CIRCUIT SENSITIVITY ANALYSIS


It is possible to build correlations between design and operating conditions and partition surface
parameters namely A, B, c and ρp to asses the performance of the separator under various scenarios
and thus to aid optimization of the separator behavior. This exercise requires a lot of data to be
collected for an operating system under various scenarios of operation.
In the absence of such correlations, one can perturb the parameters randomly within certain bounds
and study the sensitivity of separation circuit. A study for sensitivity of four possible two stage
separation circuits is done in this paper by considering constant bivariate feed distribution represented
in Table 2 and partitions surfaces defined by Eqs. (5) and (6). The selected range for random
variation of parameters is given below:
A1 & A2: –0.0010 to –0.0008; B1 & B2: –1.5 to –0.8;
C1 & C2: 0.3 to 1.3; ρ p1 & ρ p2 : 850 to 1200 kg m–3

360
SIMULATION STUDIES ON ROBUSTNESS OF MULTI-STAGE GRAVITY CIRCUIT PERFORMANCE…

1.0 1.0
Circuit 1

Recovery of density fraction to sink stream


Circuit 1
Recovery of size fraction to sink stream

0.8 0.8

0.6 0.6

3
1300 kg/m
1450 kg/m 3
0.4 0.4 1550 kg/m 3
15.750 mm 1650 kg/m 3
11.000 mm 3
1800 kg/m
7.900 mm 3
0.2 0.2 2000 kg/m
4.825 mm
1.975 mm
0.300 mm

0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6

Yield of solids to sink stream Yield of solids to sink stream


1.0
1.0
Circuit 2

Recovery of density fraction to sink stream


Circuit 2
Recovery of size fraction to sink stream

0.8 0.8

0.6 0.6

1300 kg/m 3
1450 kg/m 3
0.4 15.750 mm 0.4
1550 kg/m 3
11.000 mm
7.900 mm 1650 kg/m 3
4.825 mm 1800 kg/m 3
0.2 1.975 mm 0.2 2000 kg/m 3
0.300 mm

0.0 0.0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Yield of solids to sink stream
Yield of solids to sink stream
1.0 1.0
Circuit 3
Recovery of density fraction to sink stream

Circuit 3
Recovery of size fraction to sink stream

0.8 0.8

0.6 0.6

0.4 0.4
15.750 mm 1300 kg/m3
11.000 mm 1450 kg/m3
3
7.900 mm 1550 kg/m
0.2 4.825 mm 0.2 1650 kg/m3
1.975 mm 1800 kg/m3
0.300 mm 2000 kg/m
3

0.0 0.0
0.1 0.2 0.3 0.4 0.5 0.6 0.1 0.2 0.3 0.4 0.5 0.6

Yield of solids to sink stream Yield of solids to sink stream


1.0
1.0
Circuit 4
Recovery of density fraction to sink stream

Circuit 4
Recovery of size fraction to sink stream

0.8 0.8

0.6 0.6

15.750 mm
0.4 11.000 mm 0.4 1300 kg/m 3
7.900 mm
1450 kg/m 3
4.825 mm
1.975 mm 1550 kg/m 3
0.2 0.300 mm 1650 kg/m 3
0.2
1800 kg/m 3
2000 kg/m 3

0.0
0.2 0.4 0.6 0.8 0.0
0.2 0.4 0.6 0.8
Yield of solids to sink stream
Yield of solids to sink stream

Fig. 3: Two stage particle size recoveries and density recoveries as a function of yield of
solids to sink stream, generated by the random parameter selection for the individual
separators in a specified range.

361
B. VENKOBA RAO and S.J. GOPALAKRISHNA

Table 2: Bivariate feed distribution of the coal sample[9]


Particle size, mm
pij(d, ρ)
19 × 12.5 12.5 × 9.5 9.5 × 6.3 6.3 × 3.35 3.35 × 0.6 0.6 × 0
Float 1400 0.008791 0.010111 0.025161 0.116342 0.321257 0.178735
Particle density kg/m3

1400–1500 0.002783 0.002813 0.005342 0.014815 0.025604 0.019541


1500–1600 0.002383 0.001861 0.002587 0.007381 0.013719 0.011125
1600–1700 0.005028 0.003014 0.002425 0.005027 0.008882 0.006838
1700–1900 0.005847 0.003952 0.002884 0.005239 0.009424 0.007101
Sink 1900 0.021169 0.019250 0.017601 0.028196 0.038114 0.039660

RESULTS AND DISCUSSION


Fig. 3 shows the particle size recovery as well as particle density recovery as function of mass recovery
of particles to the sink stream for the four circuit configurations. One hundred and fifty random sets of
perturbations in partition surface parameters of the individual separators have been used to produce
these plots. These plots are similar to the ones suggested by King [2] and Tavares and King [10] for
recovery of individual density particles in the sink stream of jigs and Reichert cones. Here the
bivariate representation also permits generation of similar plots for size recovery also. Unlike
smooth curves, the scatter in the Fig. 3 is normally expected in an operating plant wherein each
operating variable is varied in a specified range.
It is seen from Fig. 3 that in all circuit configurations density recovery shows less scatter as
compared to size recovery. The density recovery patterns in circuits 1 and 3 and in circuits 2 and 4
are similar. The recovery of fines in these circuits follows the same order of their by-pass fractions
(refer to Eq. 7). The overall stability of these circuits to varying operating conditions also follows
the same order of by-pass fraction as illustrated by the range of scatter in them. Circuit 1 is the most
stable and Circuit 2 is the least stable configuration. Circuits 2 and 4 require careful control of the plant
because of the greater scatter in recovery of particles and their high sensitivity to operating conditions.

CONCLUSION
Two-stage two-product separation circuits have been discussed. The overall sharpness of separation
is improved in all the circuits. The circuits wherein the primary sink is rewashed show more stability
with regard to changes in operating and equipment design variables. Owing to the higher fluctuations
in particle recovery, circuits 2 and 4 require careful control strategies for a targeted product. The
method of random perturbation can be used to study all possible circuits with regard to their sensitivity
to operating and design variables. This technique would help in the selection of a stable circuit for
the separation of particles.

362
SIMULATION STUDIES ON ROBUSTNESS OF MULTI-STAGE GRAVITY CIRCUIT PERFORMANCE…

REFERENCES
[1] Venkoba, Rao, B., Kapur, P.C. and Rahul, K., 2003, Modeling the size-density partition surface
of dense medium separators. International Journal of Mineral Processing, 72 (1–4), p. 443.
[2] King, R.P., 2001, Modeling and Simulation of Mineral Processing Systems. Butterworth-
Heinemann, Oxford.
[3] Venkoba Rao, B., 2007, Extension of particle stratification model to incorporate particle size
effects. International Journal of Mineral Processing, 85 (1–3), p. 50.
[4] Venkoba Rao, B. and Kapur, P.C., 2008, Simulation of multi-stage gravity separation circuits
by size-density bivariate partition function. International Journal of Mineral Processing, 89
(1–4), p. 23.
[5] Venkoba Rao, B., 2007, Estimation of marginal and bivariate product distributions of gravity
concentrators using partition surface coefficients, In: National seminar on: Advanced Gravity
Separation, Editors: Singh, R., Das, A., Goswami, N.G.., NML Publication, Jamshedpur (India),
p. 96.
[6] Austin, L.G., Klimpel, R.R. and Luckie, P.T., 1984, Process Engineering of Size Reduction:
Ball milling. SME, New York.
[7] Heiskanen, K., 1993, Particle Classification. Chapman and Hall, London.
[8] Ferrara, G., Ruff, H.J. and Schena, G., 1986, Improvement of coal preparation by dynamic multi-
stage dense medium processes, 10th International Coal Preparation Congress, Edmonton,
Canada.
[9] Llewellyn, R.L., Humphryes, K.K., Leonard, J.W. and Lawrence, W.F., 1979, Dry
Concentration. In: Chapter 11, Coal Preparation, 4th Edn., ed. Leonard, J. W., SME Inc.,
New York, p. 1.
[10] Tavares, L.M. and King, R.P., 1995, A useful model for the calculation of the performance
of batch and continuous jigs, Coal Preparation, 15, p. 99.

363

You might also like