Professional Documents
Culture Documents
An Aproach To
An Aproach To
Simulation techniques are growing in importance in the Gravity concentration devices have a long history in the
mineral processing industry with respect to the design, minerals processing industry. They are used for the up-
specification and optimization of processing plant. Warren grading of ores where significant differences in specific
Spring Laboratory 0VSL) is currently developing a suite of gravity exist between the minerals of interest and the
computer models to describe the unit processes involved residual (gangue) material. Spiral concentrators are relatively
in this industry, these models being linked through mass- inexpensive devices (with no mechanical moving parts) for
balancing techniques to provide simulations of complete the separation of heavy minerals from an ore, offering a
mineral processing circuitsJ This paper is concerned with relatively high throughput per unit of installed capacity and
one class of unit process model: the modelling of gravity plant space requirement. In the spiral concentrator, the ore
concentration devices. The application of the models is (ground to at most 1 mm) is introduced, in the form of a
illustrated through reference to a particular type of slurry, at the top of a spiral channel. The particles with the
gravity concentration device, the spiral concentrator of highest densities diffuse across the bed of flowing particles
the Humphreys/GEC type. The approach was to develop towards the inner radius of the cl~annel, where they are cut
established modelling techniques into a generalized format out of the main stream (.into a concentrate product) at one
that could be readily applied to a wide range of unit pro- or several take-off ports. Sample cutters or flow diverters
cesses and to all ores likely to be processed by those devices. control how much material may be removed in this way.
Emphasis was placed on providing a model that gave im- For the GEC-type spiral concentrator, 15 take-off ports are
mediate answers to everyday problems and that could be provided equally spaced down the length of the spiral track.
directly applied by the industrial user. The separation performance for a given mineral ore is deter-
0307-904X/85/05375-05/$03.00
1985Butterworth& Co. (Publishers)Ltd. Appl. Math. Modelling, 1985, Vol. 9, October 375
I n d u s t r i a l u n i t processes: P. T u c k e r
mined both by the positioning of these cutters (machine ing on the set of model parameters (P), in providing the
settings) and by the ambient operating conditions (mass best fit. For use of the model in the predictive mode, prior
flowrates, water content of the slurry etc.). The way in specification of the parameter set (P) allows the function
which these features are accounted for, in the model, is (fp) to be directly evaluated for each particle category (i,j)
detailed below. and hence for BK to be calculated. The functional form (fp)
that can be chosen is subject to a number of constraints:
Modelling s t r a t e g y
It must be generally applicable to all ores likely to be
The modelling approach adopted was designed to provide encountered.
models of direct use at plant operational or design level. It must be robust enough to guarantee a solution.
Any data needed for input had to be both simply and The solution must provide a set of model parameters
readily obtainable. The specific model objectives were: that are unique.
To assist in the selection of a particular type of gravity These model parameters must be relatable to the physical
device for a given minerals separation problem. operating conditions.
To provide an estimate of its performance over the range When simulating measured performance, the model
of normal operating conditions and likely changes in should be relatively insensitive to measurement error
process feed. (i.e. it should smooth out rather than exactly follow this
To link with other process models in the simulation of error).
overall plant performance. In order to satisfy these constraints, it has proved essen-
To be effective the model must also be portable between tial to select relatively simple functions for the simulation,
the various types of ore likely to be encountered. A stan- the functions being limited to a maximum of four model
dard means of characterizing the ore is therefore required. parameters. The major restriction prohibiting the use of
Here, the most quantifiable and repeatable description of more complex functions, and more parameters, arises
the major material properties is given by the characteriza- through the uncertainties which can often occur in the
tion of the respective size and density distributions of the measured process data. The limitation on the number of
particles. model parameters, that can be reliably used, has required
Separation performance is characterized by means of (at least in the early stages of model development) a
coefficients of material transfer. These describe the prob- restriction on the number of variables that can be included
ability of material being transferred from the main stream in the function fp. The variables relating to the external
to the individual output flow streams. The coefficients are operating conditions are treated separately in the model.
formulated in terms of material properties (particle size and They are accounted for in terms of their producing syste-
density). The process simulation reduces essentially to the matic changes in the parameter set (P). The remaining
specification of these coefficients together with their depen- variables, which are used directly in the evaluation off/,,
dences on the internal machine settings and the external can be varied without affecting the values of P.
operating conditions.
The separation performance can be represented as:
General model structure
Ai/TKii = Brii (1)
The gravity model structure, developed according to the
where Aq is the mass flow of the ith density fraction of the
considerations listed above, is shown schematically in
jth size fraction in the process feed. TK 6 is the material
transfer coefficient describing the partition to the Kth out-
F~,ure la. The key to its operation lies in establishing the
put flow stream (BK). Experimental tests have established appropriate model parameters; once these are specified, the
that the transfer coefficient, for any given device, can be whole separation performance is described. Three paths are
adequately described by a single functional form (fp) over envisaged by which this can be achieved:
the whole range of normal operating conditions; the func- (1) The model function is fitted by regression to a set of
tion being formulated using a number of adjustable (model) transfer coefficients derived experimentally.
parameters (P): (2) Appropriate parameters are drawn from a data bank.
TKii = re(i, j, CPK, O r ) (2) Here, user judgement and experience will play a major
role.
where CPK refers to the machine setting controlling the Kth (3) Model parameters (derived via paths (1) or (2)) are per-
flowstream and OV refers to other variables (e.g. external turbed to take account of any change in operating
flow conditions). As detailed later it has proved most conditions. This can be accomplished through an
practical to deconvolute the expression for T into two auxiliary model.
parts:
The general model structure has been developed such
TKii = fp(i, 1, CPK) (3) that it is device independent. The separation mechanisms
with exploited by individual gravity separation devices (e.g.
P = function(OV) (4) hydrosizers, spirals, shaking tables etc.) and their respective
performances do, of course, differ widely. Figure lb illu-
To simulate existing plant, a least-squares fit is applied strates the way in which these differences are taken into
to the residuals between the measured (TM) and calculated account and how the general model is applied to individual
(T) transfer coefficients; i.e. minimizing: unit processes. Each unit process is represented by a module
(set of subroutines) which contains all the device-specific
Y. ZT. (TMx~i- TKq) 2
information; the major part of this is the mathematical
Kii
function (fp) used to describe the separation mechanism.
The simplex method of Nelder and Mead 2 is used, optimiz- The individual modules are designed so that they can be