OPTIMISATION
OF AN
IRON-ORE INDURATOR
R. J. Batterham (wemsen)*, J. A, Thurlby (wewsen)* and G. J. Thornton
“CSIRO Division of Chemicel Engineering, PO Box 312, Clayton, Victoria 3168,
Auswaia,
Introduction
nthe past ten years, Australis has
experienced 2 major change in the
pattern of its exports. Minerals accounted
for only 8% of total exports in 1864, but
the proportion had increased to 30% by
1974, Most of this change has been due
to the development of huge deposits of
iron ore, coal, and bauxite. The CSIRO
Division of Chemical Engineering! is @
ppartof the Minerals Research Laboratory
{0up and works in close association with
Companies which process. Australia's
The export of iron ore, one of the main
(> 6mm), which,can be charged directly to
blast furnaces, of as fines (< 6mm) which
must be converted to 8 larger size to
prevent excessive pressure drops. in
Blast furnaces. This size. up-grading
can be carried out either in sintor plants
Which are generally sited at the stealworks,
for in palitising plants preferably near the
mine ‘site, This paper _concems the
pellosing plant at Dampier operated
by Hammersey ron Pty Lid, which i
‘one ofthe few targe capacity processes in
‘Australia, having’ throughput. much
higher than” most Australian chemical
processes.
sections of the mineral trade, is from the
Pilbara region of North West Australia.
which includes two of the world’s largest
mines, The product is sold as lump ore
lle
The Pe
ising Plant
Details of this plant have been given
elsewhere?
Ore (< 6mm) is blended and
=]
Fig 1—The induration machine
Attar graduating from Molbourne University (PhD) in 1968, Robin Battezham (left)
worked at the ICI Central Instrument Resoarch Laboratories at Reading before
‘moving to CSIRO. His main research interest Is
Imineral processing plants.
A graduato of Melbourne Univesity, BE (Chem) in 1962 and ME in 1976, J. A,
Thurby (contre) sent nine years with Altona Petrochemical Company and Esso Aus-
tralia before joining CSIRO. His current interests are process modeling end
Gu. Thornton (right) spont several years inthe aircraft industry before joining CSIRO.
Ho isan electronics engineer with nterestsin process control and systems engineering.
the contol of large-scale
‘THE CHEMICAL ENGINEER, SEPTEMBER 1977
‘ground to a very fine size to allow balling
into moist “green” pellets which have poor
hendling propertios and must be heat
Iweated (indurated) at approximately
130°C to develop adequate strength
Much of the bonding in the fred pellets
results from reactions with amounts of
lime which are added. et the grinding
stage to ensure an adequate dispersion.
‘The firing process is complex?, involving
the formation of ime-hematite-sicate
‘compounds which facilitate diffusion and
{grain growth in the hematite particles ta
form a hard pellet on cooling, with the
ability to resist breakage in transport and
handling
The induration machine used at Dampier
is shown ciagramatically in Fig 1. The
moving grate is protected from excessive
(gas temperatures by @ layer of previously
fired pellets refered to. as the “hearth
layer”. Green pellets ae sized by screening,
placed on top of the hearth layer, and
vey _move through the machine are
successively heated gently. to. remove
free water added in balling (33% of the
machine), heated at @ higher temperature
(600°C) to remove the chemically bound
water contained in goethite in the ore
(5% of the machine), and then fred
at 1300°C for sutfciont time to allow the
Strength to develop (28% of the machine)
Finally, the pellets are cooled to reasonable
handling temperatures.
The gas flow pattern shown in Fig 1
lustrates the methods used to minimise
energy consumption, which is about 1
tonne of fue! oil per 50 tonnes of product
The machine ‘operates at maximum
opacity and there is considerable incen-
tive to increase the tonnage produced at an
‘acceptable quality of pele.
The process is highly interactive, for
example, alteration of the gas flows in the
firing section of the machine will affect
the gas conditions in the drying sections.
This wil alter the bed temperatures in the
drying zones, which will cause further
changes as the bed moves into the firing
zone. Such gross. interactions are
‘obvious to the operators but many process
changes have subtle effects which opera
{ore find dificult to assess and rectify. The
‘machine also has significant and vated
ttansport delays not normally met in the
process industries: the ore residence time
'S ono hour, yet the residence time of the
{gas from ‘cooling through fing and
through drying is only ten seconds. For
control, the problem of delays ie further
compounded by the extra delays involved
in measuring the product quality. Samples
fof fred pallet are analysed every two
hhours and may pertain to operating
Conditions up to thrae hours previously.
Thus the general problem of Improving
the economie performance ofthis process
629hhas two main aspects: (1) elimination or
‘amelioration of the effects of delays in the
process, espacally on product quality, anc
(2) elimination ofthe effects of interactions
in the process that are too subtle for
‘operators to detect.
"The approach has been to develop
‘an analytical model of the process to
provide information “on the process
Interactions and on the effect of various
‘operating strategies on machine. per-
formance. The economics justify @ major
cetfor, and several man-years have beon
expended to date, particulary on the
‘experimental. side, 98 part of 2 joint
esearch programmo by Hammersey Iron
Pty Ltd and the CSIRO Division of Chem
cal Engineering,
The model building has raliod heavily
fon results from pot-grate_ simulators
laboratory-scale machines. which uso 2
small fixed bed of pellets which is
Subjected to gas flows and temperatures
similar to those on the plant. In eect, this
provides the equivalent of 2 slug of
material moving. through the. fullsca
Machine." The method has been used
fextonsively to. give information on
ja drying rates? and the effects of
time and temperature on the quality of
pollet produced. The resuitant data can
‘also. be checked by instrumenting one
Section of the real’ machine bed with
thermocouples and following its passage
through the machine. AS it is not easy
to measure the gas flows on the machine,
the related bed pressure drops. are
measured, and, as 2 consequence, the
comparison between pilot scale and the
plant can never be complete.
The development of steady-state
made! has reached the stage where the
ftfect of various operating parameters
‘can be defined and used to dafine
fecanomic operating conditions. Even
tually, this work will be extended to
‘cover the dynamics of the machine and
will be used t0 determine whether optimal
Contral polices exist and the extent to
Which effects of ‘interactions and. time
‘olays can be minimised.
Optimising Controt
To satisfy immediate plant requirements,
fan optimising controller has been deve
Toped which is partly based on an
‘operational model of the process. The
‘model uses statements of logical relation
Shipe betweon process variables (6g. 92s
temperatures, pressure drops and flows)
and the plant states (e.g. product quality
at various depths in the bed, drying rates
and bod tomperatures). The technique is
Suited to. processes. where there
frequent variations in process. con.
Gitions and hence inthe process optimum.
Ite. main advantage is that the control
action specified fora given set of input is
Unique, and a plant performance at least
fequivalent to that obtained by the best
process operator is possible.
Conventional optimising algorithms
using ill climbing techniques incon:
junction with an analytical model cannot
be used in this case because of the
complexity of the process. Evolutionary
‘optimising methods are also inappropriate
a it is not possible to maintain machine
Conditions steady enough to observe the
20
effect of chenges to process variables.
The section of the control scheme
developed forthe firing zones uses an on
line estimate of the peliet quality and
thereby avoids the main delay in the
‘process, whioh is the measurement of
pellet quality. Detals of the optimising
‘Controller have been published elsewhere"
‘The present paper concentrates on the full
analytical model of the machine and some
fof tho rosults of computer simulations
With a brief description of the on-lino
‘Quality preditor which uses 9 smaller, but
Similar, heat transfer model for the firing
On-Line Quality Predictor (Firing
Zone Operation)
To predict the temperature profiles of
pellets st various depths in the bed as the
pellets move through the machine, @ heat
ttansfer model has been developed which
Uses the measured hood and winsbox
temperatures (i.e. the temperature of
gases entering and leaving the bed)
{ogether with the observed pressure
drops across the bed. At any given
location in the machine (ie. at any given
‘windbox) the bed is divided into a humber
‘of horizontal elements and the gas and
pellet tomperstures ate calculated for each
element. The physical size of each element
fs Kept small enough to keep the model
equations linear. (As only the firing
section contributes to pellet quality. the
‘drying section is ignored)
No allowance has been made for such
second order effects as axial dispersion in
the gas, lateral mixing ofthe ges. channel-
ling’ through the bed, and conduction
through the bed. Radiation has been
included ‘but generally has lite sign
ficane® except for the top layer which is
‘exposed tothe oil flame in the hood above
the bed,
‘Various other restrictions apply. Thus
‘the model assumes a uniform pect size, no
temperature distribution within the pellets,
‘and. 2 uniform bed height along tho
machine (ie. no bed shrinkage in the
firing zone)
Tits simplest form, the model est-
mates the rate of heat transfor from tho
{998 10 the pellets and the heat balance
Betwoon tho gas and the pellets. The
rate equations aro. simplified by using
Several lumped parameters; viz constants,
Known functions of temperature. which
‘compensate for changing physical prop.
fenios, and factors which depend on the
bed geometry and which relate the
observed pressure drops to the actual
‘unmeasured gas mass flows. These lumped
parameters are updated on-line so that
Predicted gas tomperatures leaving the
bed egree with plant observation.
Having calculated the pellet tempera-
tutes throughout the firing zone, @ state
‘equation is used to estimate the resulting
pellet quality. This equation predicts the
Influence of any arbitrary time-tempera-
ture profile on the pellet quality produced.
Each of the on-line models has been
tested extensively against data generated
‘on pot-grate simulators where the temper-
Stures at various levels inthe bed. are
known, together with the resulting pellet
quality, Figure 2 shows the predicted and
observed bed temperatures for tho fring
zone for atypical run. (The cooling zone
that follows is essentially 2 miror image)
I'has been shown that the agreement is
better than that obtained using simplar
models! which donot. allow. for the
effect of temperature on physical proper-
Figure 2 also shows how pellets at the
top of the bed reach @ higher firing
temperature than pellets. lower down
and” that. this higher tempersture is,
maintained longer. As a consequence
there is a range of qualities produced in
the bed with pellets at tho bottom
being of lowest quality. Thus the pellets
‘which form the “hearth layer” must come
from previously fred pellets of adequate
quality, 8s the bottom of tho bed shows
Title or no increase in quality in one
passage through the machine.
’As mentioned earlier, the contol scheme
based on plant operational logic and the
folatively ‘simple on-line models is 8
method of optimising the process which
oes not involve the laborious step of
Goveloping a full analytical model of the
process.
Fig 2—Measured and predicted temperature profiles forthe firing section
‘THE CHEMICAL ENGINEER, SEPTEMBER 1977Full Plant Simulation (Drying,
‘and Cooling),
“The objective of this portion of the
research has been to provide an off-line
‘model that can be ‘used for optimal
Control studies and ‘also to. perform
Strategic ‘model-based “plant tials” 10
etermine the optimum ‘operating con-
Gitions, og. optimum ratio of bed and
hearth layer heights versus grate spoc
freon all size and bed. porosity inter
Actions, the importance of fan capacitios,
‘ate. Those trials are cheaper and can cover
wider conditions than those acceptable on
the actual plant, without the risk of lost
production.
‘The model covers the entire machine
(atving, fring, and cooling) and includes
radiation effects, influence of free water,
ehydration of gosthite and othar hygrated
minerals, limesand calcination and fring
reactions. In the drying section, the
predicted drying rates are compared with
the maximum rates that pellets can Sus:
tain without exploding (spalling) due to
the pressure of steam within the pellets,
This’ prediction of spalling is essential in
‘any realistic. simulation as, on the plant
Seale, contol of spaling can dominate the
‘whole operation
Asin the on-line model, certain
assumptions are made, viz uniform pellet
Size, bed geometry, bed height. and gas
distribution across the bed. The grate and
Understucture are considered as 8 number
fof elements, with heat transfer within an
flement by convection and where approp-
fiate by radiation; heat conducted
between the elements is also taken into
‘account
Radiation to and through the bed and
conduction through the bed are included
inthe model but do not have a great effect
fon pellet temperature profiles. The
Simulation is handled. in two. separate
but related computer programs. "In the
first, caleulations are made using a
specified set of gas temperatures and
compositions entering the bed together
‘with specified mase flows oF pressure
‘ops. The predicted outlet ges conditions
{rom each bed element are blended to
calculate windbox flows, enthalpies, and
Compositions. This. program allows
‘xploration of effects of the main variabios
‘on spalling and pellet quality, e.g. tho
effects of bad inlet temperatures, flow
rates, pellet moistures and. size, bed
porosity, and grate speed. However, it does
fot calculate overall heat and mass
balances or give any indication of the
feasibility of the chosen conditions.
Results of this open loop simulation
show satisfactory agreement with data
{generated in pot-grate simulators, and, as
far as can be observed, with the full sal
plant. Figures 2 and 3, which cover the
firing and drying sections respectively,
compere experimental and simulated
results
The second computer program is for
recycle flow and. incorporates. the fist
program for calculation of pellet tempera
{ures and bed outlet temperatures. Opera
ofthis program is best understoad by
referring to Fig 1. On completion of the
bed. calculations, the gas flows leaving
tach bed element are blended to give flow
fates, enthalpies and. temperatures of
the stream passing from the first Cooling
Table—The Ranges used inthe Simulations
Variable Units Type. Lowest «Base Highest
value state value
Tonnage Tonnes/in Fixed = 400 ea
Bed Voidage mi/m> Independent 31 33 35
Pellet diameter m ‘ont ‘ota7_— “ots
Depth of hearth layer m 05, 10 18
Depthofareen pellets m 4 48 58
Total bed depth m Dependent 46 56 66
Gratespeed m/min 164 200-225
0 oor
peo must rewrenarune
ile (HAR)
Fig Temperature profiles and spalling effect in the drying zones
Fig 4—Caleulated parameters in the bed at various positions along the machine, The
numbers refer to the windboxes
Zone 10 the Preheat/Fiing Zone, the
‘steams to Windbox Recuperation Fens NO
4 and No 2, and the stream to the Hood
Fecuperation Fan. ‘The airleeds required
{o attain specified fan discharge tompor-
‘atures are also calculated, From this
information, the tempering air requirements
for the drying zones can be calculated
together with the fuel and atomising ai
flows in 2ll other parts of the machine.
With these new gas compositions, the
bed calculations are repeated and the
‘whole process reiterated until a satisfac-
tory convergence occurs. Such physical
limitations a8 the impossibility of negative
air-bleeds are imposed,
‘Some second order effects or
for in the model, These include
above and below the bed together with
hood sealing air requirements, heat
losses from all ductwork, and by-passing
fof gases from the Updraught Drying
Zone to the first Down-draught Drying
‘THE CHEMICAL ENGINEER, SEPTEMBER 1977
Zone and from the fist Cooling Zone
to the last few windboxes of the Firing
Zone.
Simulation Results
‘The Tablelists the base state andtherange
‘of the variables investigated. This sot of
Simulations was performed at a constant
tonnage which forces two of the six
variables to be dependent, the other four
being independent. The main invest-
gations have been to vary the bed
Porosity, the poet diameter, and. the
heights of the hearth layer and the green
pellets
Figure 4, summarising a portion of the
‘output, shows how selected gas proper-
fies’ vary down through the bed at
various. windbox positions along. the
‘machine forthe base state
Results for the drying section are
shown in Fig 6 and for the overall plant
economics in Fig 6. The abrasion index is
6st