Вы находитесь на странице: 1из 3

Proceedings of Fourth

Australan Compter

Conference
Adelaide, South Australia, 1969

The Simulation of Mineral


Processing Machinery
By W. J. Whiten
Department of Mining and Metallurgical Engineering
University of Queensland, Brisbane
ABSTRACT: The use of computers by a University of Queensland research project for steady
state and unsteady state simulation of mineral treatment processes is descbed. Several systems
of Fortran programmes have been used for this work. The design, advantages and disadvantages of these programmes are discussed.
WORDS: Mineral dressing, mineral processing, continuous simulation, digital simulation,
steady state simulation, simulation languages.
COMPUTING REVIEWS CATEGORIES: 3.22, 3.29.

INTRODUCTION
research group at the Uiversity of Queeslad supported by
the Australia Mineral Idustries Research Associatio ad
Mout Isa Mies Limited, has bee workig for several years
the simulatio ad cotrol of mieral treatmet processes.
The object of the research work is to develop techniques for
the optimisatio ad automatic cotrol of these processes. Up
the ed of 1967 the work was mainly griding (the
breaking of ore to a fine powder) ad classificatio ( the separatio of ore into fine and coarse size fractions) . This work
has w bee exteded to iclude froth flotatio ( the cocentra
tion of valuable mieral grains using surface iteraction with
bubbles). Simulatio of both steady state ad usteady state is
being studied. Post-graduate studets associated with the project are workig the crushig of ore, flow through storage
bins, ad wet cocetratio of beach sands.
The followig equipmet has either bee simulated a
digital computer or is being aalysed for simulatio:
1. Commiutio (i.e. size reduction of ore lumps)
Rod mills
Ball mills
Coe Crushers
2. Classificatio (i.e. separatig coarse particles from fie)

Hydrocycloes

Vibratig Screes
Cocetratio (i.e. removal of
Flotatio Cells
Tray Gravity Cocetrators

Up to the preset time the


225 computer has bee used.

also
waste from ore)

Unversity

of

Queeslad's

GE

DEVELOPMENT OF MODELS
The models are developed ad verified wheever possible with
data from full scale idustrial equipment usually supplemeted
by data from pilot plat equipment. We have foud that the
most satisfactory models are developed from theory with parameters being calculated to fit the experimental data.
Ofte it has t bee possible to develop a suitable model
from theory ad these cases empirical relations are derived
from experimetal data. Stepwise multiple liear regressio has
bee used extesively for the reductio of data but it has t
proved very satisfactory. Some other methods of data reductio
are to be evaluated.
MODELS USED
the type of models being used for simulatio, the
two most used steady state models are described briefly:

idicate

fractio

S is a diagoal matrix gg selectio rates for the breakage


of each size fraction
c is the ball mill costat.
The toage ad water fed to the mill are assumed the same
the product. Whe is o-itegral p s calculated by
lear iterpolatio from the itegers earest to .
2. The Hydrocycloe (a separator of fie ad coarse sizes of
ore particles).
The beha viour of the hydrocyclone is described by a
reduced efficiecy curve, a split size d 50, ad the amout of
water goig with the fie ad coarse products.
The reduced efficiecy curve is calculated from
e 1130 -
()

Rake Classifiers
Sieve Beds
3.

1. The Ball Mill (this machine breaks the ore particles).


The discharge p (a vector cotaiing percets a rage of
size fractios) is calculated from
= c/r
= ( S + I - S)nf
where
fs the feed size distributio vector
r is the feed rate
is the breakage matrix, i.e. a descriptio of the average
size dstrbutio produced by the breakage of each size

Loge (d 50) = .886- .719u + .219 - .0443W 0 + b 3


W = 1.1 W 1 + b2 - lOu
p = Q,2/(k z1 vz(l-S,). 24)
Where
is the diameter of the overflow (the fie size exit)
u is the diameter of the uderflow (the coarse size exit)
W 1 the toage of water the feed
W 0 the toage of water the overflow
S1 the fractio solids the feed
Q, the volume of the feed
the operatig pressure
k 1 k 2 k 3 are parameters.
The refereces give the models
worked examples.

greater detail

ad

also

USAGE OF MODELS
The models allow the calculatio of the complete descriptio of
the product provided the feed description ad the properties of
the machie (i.e. the model parameters ) are kw. The simulatio of simple processig plats without closed circuits is
doe by following the flow of .ore through the plat calculatig the product of each uit tur.

577

In the case of closed circuits which are the more usual, an


iterative procedure is used. An initial approximate description
( often zero) of the circulating or returned load ( or loads) is
used to allow the calculation to proceed until better ( we
hope) values are obtained to replace the initial approximation.
The physical background to the iterations indicates that they
will converge. Normally 4 to 20 iterations are required for
convergence.
Optimisation and design of control systems have up to now
been done by working from the printed results of simulation.
Manual procedures had been adopted here due to the limited
amount of core available, the difficult nature of some restraints
and the requirement to produce results quickly. This has provided some useful jnsight into the behaviour of the models but
now that a larger computer is available automatic optimisation
appears more economic.
SIMULATION PROGRAMMES
Firstly we look at some of the factors affecting the design of
the simulation programmes. These are:
1. Data describing one machine or ore flow may have a
complicated structure.
e.g. The description of an ore flow may require the
following:
The amount in each size fraction.
The sizes dividing the size fractions.
The percents of the vaous ore components in each
size fraction.
The specific gravities of the ore components.
The amount of water in the ore.
Also the amounts of various reagents and composite
particles may be required.
2. Between simulation runs usually only a few of the values
describing the plant are changed. Which values change
is usually unpredictable. Thus an input system that
accepts changes to the data is desirable.
3. Turn around times of 2-3 runs per day have been found
not unsatisfactory. This is somewhat faster (at present)
than the University of Queensland's punching service
hence it has been required that the preparation of simulation runs involves a minimum of data preparation.
4. Modification and alterations to programmes are often
required as new models are developed.
5. The programmes should be simple enough for use by
industry and students. When the simulation programmes
have been 'tested and vefied they are made available to
the companies supporting the research and are used for
teaching in the niversity.

The following criteria-. have been used as a guide for the design
of programmes:
1. Detail should be required only in the sections of programme where it is used.
This ensures that programmes do not contain unnecessary complication and changes can be made
easily.
2. change to one part of the system should not affect
unrelated parts.
i.e. the programmes should consist as much as possible
of completely independent subsections.
3. Programmes should be of general purpose design.
provide greater system utilisation and reliability.
reliability.
4. The data should be simple to prepare and check.
provide greater system utilisation and reliability.
one system of programmes has been written (nor appears
feasible) to satisfy all of the above. The choice of programming languages was between assembly language (GE 225
GAP) and Fortran 11 with Fortran IV coming available later.
Fortran has been used except for a few assembly language
subroutines for list handling. The compromises that have been
developed are descbed in A-D below.
.

Fortran Subroutines
Initially a sees of subroutines for simulating the various
machines were wtten in Fortran 11. lnput, circuit description,
convergence tests, and output were all coded into the Fortran
main programme.

578

These programmes were quite efficient in their use of storage


and machine time.
The disadvantages were:
1. The data decks were long and rather sensitive to errors.
2. The main programmes contained too much information
and thus were not sufficiently easy to change.
3. Maintaining the programmes and subprogrammes proved
a problem. Main programmes were required for several
trivially different simulations. For each of these programmes instructions for data preparation had to be kept
and also when subprogrammes were updated many card
decks had to be altered.
.

General!sed Simulation Programme


increase the efficiency of the simulation system a programme was written to read the circuit descption and accept
changes to the circuit description and the data. This programme was written in Fortran 11 with the exception of a subprogramme to preread characters from the card input. The
simulation subprogrammes written previously were incorporated into this system. The features of this system are:
1. The circuit to be simulated is described by a series of
simulation commands, e.g.
*01 BALL MILL 10FTX12F FEED PRODUCT
This description is converted to an integer code and then
interpreted.
2. Twelve character data names are defined by their use in
the circuit descriptions. These names are used to identify
input and output. The internal data areas are defined
dynamically their types being found from implicit information in the circuit descption.
3. The input data is checked for errors. U an error is found
it is noted then the programme continues if possible.
4. Each simulation is started by a *EXECUTE data card.
After each simulation any parts of the data, the circuit
descption
and dimensioning information may be
changed.
5. The output is completely automatic and follows the order
given in the circuit description.
6. Iterative loops are abandoned should any of the machines
being simulated become grossly overloaded.

This programme has proved very efficient and convenient to


use. The simplicity of the circuit descption and the lack of a
fixed order for the data means that a sees of simulation runs
can be made quickly with little chance of error. The extra flexbility provided by this programme rapidly paid for the extra
development costs.
The main deficiencies of this system are:
1. The main programme is at present too long for easy and
cheap alteration the GE 225.
2. V ery little core storage ( G 225) is left for any expansion of the system.
3. Facilities beyond predicting output given input are not
available.
Experfmental List Structure Simulation
Next it became necessary to provide for more complex data
and unsteady state simulation. An experimental system was
written to evaluate the use of data structures. Fortran IV was
used as the basic language with several assembly language subprogrammes to give efficient handling of the data structures
and subroutines with a variable number of arguments.
The basic data structure consists of homogeneous vectors
containing one of real values, integer values, or pointers to
other vectors. Each vector has a header element containing a
type code and number of elements.
The main features of this system are:
1. Each data structure
( a) ls referenced by a single integer pointer
(b) Cares an alphabetic name used for input and
output
( c) Cares input and output directions including formats
(d) ls on the list of all data structures.
C.

2. Data structures are defined by calling fortran functions


giving the structures alphabetic name as an argument.
These functions build the structure and provide a pointer
to it as the function value.
3. Input is preceded by either the data structure name or
codes for forming a vector referenced from a COMMON
pointer region.
4. Fortran subroutines are used for the simulation. Only one
argument is required for each data structure used by the
subprogramme.
5. Convergence tests are performed with the aid of a functon that calculates the relative change in its argument
dung the last iteration.
6. Output is effected by giving one or more data pointers to
the ouput subroutine. New pages are automatically provided, headed and numbered at suitable points in the
output. The single line calling for output may cause the
pnting of several pages using perhaps 20 different formats th~ order of which may vary with the size of the
data.
7. Continuous (unsteady state) simulation is accomplished
with the aid of a function that gives a eode to determine
the next operation and advances time as necessary.
8. Debug facilities provided with the data structures are:
(a) trace of vector storage allocation and retrieval
operations.
(b) trace of the data structure addressing operation.
(c) dump routine for the data structure memory.
The following is an example of a main programme to simulate the behaviour of a ball mill:
CALL
IFEED
LPULP ( 12 FEED
IPROD
LPULP ( 12 PRODUCT
= LBALLM (12 BALL MILL
10
CALL READ
CALL BALLM (IBM, IFEED, IPROD
CALL PRINT (, IFEED, IPROD

=
=

ao

END
Not much more than the basic description of the problem is
required in the main programme. Detail not required here is
kept within the appropriate subprograms.
This system provides a simple method of describing the
simulation problem and still allows the full generality of fortran to be used when required. The data can take a far more
complicated form than can be used with any degree of ease
with the more usual techniques in fortran. The amount of storage used to implement this system is considered to be quite
reasonable.
Due to the amount of core (5) available for programmes,
segmentation is usually required. The use of a COMMON
system for data communication is now not considered satisfactory. Some penalty is paid in the running times required for
simulation.
Further development requires a larger computer.
D. Fortran IV Subprogrammes
provide programmes for student use, use outside of the
University and rapid utilisation of the new PDP 10 computer a
system of Fortran IV subroutines has recently been written.

These programmes have the following characteristics:


1. provide a uniform system, subprogrammes for all
normal simulation operations have been written. These
include simulation of each process, the compason for
convergence tests, and also input and output of each data
type.
2. The programmes have been designed for use as il subsystem in other programmes such as optimisation.
3. lnput is identified by order only. Output is identified by a
12 character name provided as an argument in the output
subroutine calling sequence.
4. COMMON store has not been used for communication
between the basic subprogrammes. Instead greater fiexibility is obtained by transmitting an aay of often
required information as an argument.
While this system has the advantage of using only simple
programming techniques, it has the following deficiencies:
1. Often several subprogramme arguments are required to
describe one piece of equipment. When more complex
data is required the number of arguments becomes
excessive.
2. The data are not named thus making checking of and
alterations to data harder. Further the order of data is
predetermined by programme.
3. The wting of input and output sections requires one line
for each machine of pulp stream which is long compared
with c.
4. Although the use of optimsaton routines is relatively
simple the process is not as tidy as would be desired.
5. Care has to be taken in accounting for system
changes.
CONCLUSION
Fortran has been used as a simulation language. Various
methods of overcoming the deficiencies in fortran for this usage
have been described. The programmes described in this paper
have proved of practical use for the optimisation and control
of mineral processing machinery. The simulation of a grinding
section at Mount lsa Mines lead to the reaangement of the
section which in increased the throughput from 115 tph to 125
tph. An automatic control system designed by simulation
further increased throughput to 130 tph. These techniques are
now being applied to several other plants.
ACNOWLEDGEMENTS

Are due to the Australian Mineral Industries Research Association, Mount Isa Mines, the staff of the University of Queensland Computer Centre, and in particular Dr. . J. Lynch.
REFERENCES
DRAPER, . , LYNCH, . J., W, W. J. (1967): "Developing
the Optirnum Performance of a Mti-stage Grinding Circuit."
Transactions of the lnstitution of Mining and Metallurgy, Vol. 76, p.
C169.
DREDGE, ., LYNCH, . J. (1968): "Course Notes-Second Residential
School of Mineral Processing."
ELLY, J. R., LYNCH, . J., RAO, . C., W, W. J. (1967):
" Analysis of the Performance of a Ball Mill-Rake Classifier
Cornminution Circuit." Aust. ... Proceedings, . 224, p. 9.
LYNCH, . J., RAO, . C., W, W. J. (1967): "Technica1 Note
on On-Stream Sizing Analysis in Closed Grinding Circuits", Aust.
... Proceedings, . 223, p. 71.

The Griffin Press, Adelaide

579

Вам также может понравиться