A Dynamic Mathematical Model of The Complete Grate/Kiln Iron-Ore Pellet Induration Process
A Dynamic Mathematical Model of The Complete Grate/Kiln Iron-Ore Pellet Induration Process
A Dynamic Mathematical Model of The Complete Grate/Kiln Iron-Ore Pellet Induration Process
Induration (drying and hardening) of iron ore pellets is an energy-intensive feed preparation step for
both the blast furnace and direct reduction routes to iron. It is commonly carded out by a 'grate/kiln'
process. A mathematical model of the process is described, in which mechanistic models of the grate
furnace, kiln, and cooler are linked to enable simulation of the entire process. The model includes
equations for the gas stream pressure balances and process controller responses, and provides dynamic
solutions. The validation of the model for an operating plant is discussed and steady-state solutions
are compared with data from the plant.
I. INTRODUCTION moist balled ore, a rotary kiln, and an annular moving grate
cooler. Figure 1 depicts the process for a two-pass grate
A S with many other metallurgical processes, induration furnace, as used at Whyalla. Green pellets are deposited
process developments have largely been through direct plant as a 180 mm deep bed on the grate and are subjected to
observations coupled with laboratory testwork on specific downdraft drying by gases at 350 ~ before passing into
aspects. Routine measurement of some key plant variables the preheat furnace. Here, drying is completed; partial de-
is impractical and steady-state operation is hard to achieve composition of carbonate fluxes and oxidation of magnetite,
due to instabilities in green feed preparation, 1 high solid mill scale, and supplementary fuel such as coke breeze in
residence times, and recycling of gases. Consequently, anal- the pellets, also occurs. The mean pellet temperature is
ysis of plant behavior is often difficult and inconclusive. increased to about 1000 ~ ensuring sufficient intergranular
Physical simulation gives valuable information on ore pro- bridging for the pellets to withstand tumbling in the kiln.
cessing characteristics and on specific problems, but has
limitations as a means of representing a whole process. 2'3 For acid pellets considered in the present work, the pellets
are heated to about 1330 ~ in the gas or oil-fired kiln. The
In their recent review of fundamental aspects of sintering
and induration processes, Wynnyckyj and Batterham4 con- two-stage cooler subsequently recovers most of their sensi-
cluded that future developments can and should proceed ble heat, either as high-temperature secondary air for the
kiln or in a medium-temperature air stream used in the
from the theoretical base which is now well established. A
drying furnace. Air flow through the pellet bed in the cooler
number of models of the complete Lurgi straight-grate pro-
is updraft, provided by fans discharging into two separate
cess have been validated to varying degrees against oper-
ating plants. 5'6'7 Plant-validated shaft process models have windboxes.
also been reported, s However, there is only one published Two 'preheat' fans draw kiln offgas through the bed in
model of the complete Allis-Chalmers grate-kiln process, the preheat furnace. They deliver this, together with air out
that of Young, Cross, and Gibson. 9 The three main equip- of the 'heat recuperation' duct from the second stage of
ment items (grate furnace, rotary kiln, and product cooler) the cooler, to the hood of the drying furnace. Gases are
are mathematically linked in this steady state model, but drawn through the bed in the drying furnace by two 'waste
influences on gas flowrates were not included. Without gas' fans.
these, and validation against an operating plant, only ten-
tative conclusions could be drawn from simulations.
This paper describes a dynamic model of the grate/kiln III. MATHEMATICAL MODEL OF PROCESS
process, which includes gas flow resistances and character-
istics of the process fans and control systems. Its devel- Dynamic models have been developed for each of the
opment was undertaken by CSIRO in collaboration with the three main plant items--the grate furnace, kiln, and cooler.
B.H.P. Steel International Group-Long Products Division, These are solved repeatedly in sequence. Boundary con-
to improve the efficiency of the B.H.P. plant at Whyalla, ditions for each model are provided by previous solutions
South Australia. Results of simulations for recorded plant of the other models, full account being taken of the gas
conditions are presented. pressure/flow relationships within the process.
Companion papers 1~ report results of studies aimed
principally at reducing energy consumption. A. Equationsfor Grate Furnace and Cooler
Assumptions made in modeling these items are as follows:
II. GRATE/KILN PELLET INDURATION (i) The bed is initially homogeneous, the gas flow does not
vary across the bed, and heat losses from the sides of the bed
The Allis-Chalmers grate/kiln process employs a fur- are negligible. It follows that all state variables are constant
nace with a moving grate for drying and preheating the across the bed.
(ii) The velocity of the gas is much higher than the grate
J.A. THURLBY is Principal Research Scientist, Division of Mineral
Engineering, Commonwealth Scientific and Industrial Research Organi- speed.
sation, Clayton, Victoria, Australia. (iii) Radiant heat transfer, and conduction within the bed,
Manuscript submitted June 24, 1986. can be ignored. 6
METALLURGICALTRANSACTIONSB VOLUME19B,FEBRUARY1988--103
PRIMARY PRIMARY
AIR OIL GAS AIR
DRY,NG I PREHEAT ~ / /
FURNACE I FURNACE I _ I ~ != i I
GREEN
BALLS
~ l-l-l l l l l l l ~- -I-/- l l l l l l /
-- / ~ . KILN l " - ~ " PARALLEL FLOW DUCT
,,.
i", li J , " "
] 11st STAGE [2rid STAGE I
TO SCRUBBERS
& STACK
(iv) Pellets are initially homogeneous. Concentrations of pellet constituents are given directly by:
(v) Temperature gradients developed within pellets can be
ignored. OXj _ R i [6]
(vi) Heat liberated (or required) by reactions within pellets O0 Pb
is absorbed (or supplied) solely by the solids. and heats of reaction (excluding condensation and dry-
(vii) Heat liberated on condensation of water is absorbed ing) by:
solely by the gas.
The cooler model is the same as the grate furnace model, qcr = E Rj AHrj [71
except for the omission of equations for drying and calci-
nation of limestone and dolomite. Also, as cooling is up- Of the rate terms required for solution of the differential
draft, heat transfer to and from the grate is negligible. Eqs. [1] to [6], heat transfer by convection is given by:
At any point z m below the top of the bed, the gas tem-
perature gradient is given by:* qcv = h A (Tg - Tp) [8]
* A list of symbols is given at the end of the paper. The remaining terms are discussed below. Calculation of
heat and mass transfer coefficients, and reaction rate con-
OTg _ qcv + qca [1] stants, is discussed in Appendix I.
Oz GCg 1. Drying and condensation
Throughout the course of drying of single pellets under
and the rate of change of the solids temperature by: constant conditions in the industrial range, it has been
OTs _ qcv + qcr - qdr [2] found 13 that the heat used in drying remains essentially at
a fixed fraction of the heat transferred to the pellet. This
O0 pbCs
fraction, fd, depends primarily on the gas temperature and
The pressure gradient is given by the Ergun Eq. [12], hence changes continuously during the nonisothermal dry-
i.e.: ing in a bed. For gas temperatures below 100 ~ it has a
value close to unity. In this temperature range the humidity
0__p_P= 150 (1 - 6)2~g G 1.75 (1 -- e)G2 [3] of the gas is usually quite high and diffusion of water vapor
Oz ~2 d2 83pg ~) de 3pg limits the drying rate. Thus,
As the gas passes through the bed, the total flowrate and Rdr = fdqcv for Tg >-- 100 ~ [9a]
the concentration of individual species change (due to AHdr
reaction, drying, or condensation), according to Eqs. [4] or
and [5].
Rd,.
=(\ mat- )
_H__~atH . fdqcv for Tg < 100 ~
OG AHar
-- E e i [4]
0z [9b]
GAS FILM
COKE C O M B O S T , O N
,,TE.FAOE "%
MAG.ET,TE
OXIDATION
INTERFACE /
n-
I-
Z
LU
0
Z
o
o
r
o
LAYER (1) Fe 2 0 3
LAYER (2) Fe 2 03+coke ( P c kg m-lpellet)
CORE Fe2 0 3 + c ~ Pm kg m -1 pellet)
Fig. 2--Simultaneouscoke combustionand magnetiteoxidation.
106--VOLUME19B,FEBRUARY1988 METALLURGICALTRANSACTIONSB
G A S TO
PREHEAT
C. Calculation of Boundary Conditions
FURNACE
ROTATION Most of the boundary conditions (i.e., solid feed and inlet
gas temperatures and compositions, and the inlet gas flow-
rates) for the individual models depend on the process as a
OED~ARY whole and change until steady-state operation is attained.
BURDEN FUEL & The solution at each time-step also requires initial values for
WALL FEED
MARY AIR all state variables pertaining to the solids. To commence a
FROM
FURPREHEAT
!R~E~: AT simulation, the latter and boundary values for the first model
PRODUCT (the cooler) are usually provided by a computer file from a
TO COOLER
previous simulation. Subsequent boundary condition calcu-
Fig. 4 - - Schematic d i a g r a m o f kiln.
lations are discussed elsewhere. H
IV. VALIDATION OF T H E M O D E L
duction through the wall, heat losses from the kiln shell, and
transfer within the burden by mixing. For radiation, the The validity of a whole-process model such as this needs
contribution of all emissions and multiple reflections are to be assessed relative to the task for which the model is
included. This is a significant advance on most kiln models, used. In this case validation has been mainly directed toward
which ignore certain radiationpaths, e.g., reradiation from studies on steady-state optimization and process design
the burden to the wall, 15 or consider
~"
only the first few reflec- modifications, rather than dynamic control.
tions and omit the flame. 16 For the Whyalla plant, radiation and convective heat
Rate equations are integrated in finite difference form, the losses from ducts and main equipment items were de-
kiln being typically considered as 50 cylindrical zones. termined by calculating conduction rates for measured in-
With the assumption of pseudo steady-state operation of ternal and surface temperatures. This gave a set of base
each zone, the heat fluxes at the inner surface of the wall and values (Table I) which are adjusted during a simulation in
at the kiln shell are equal, and the inner wall temperature accordance with calculated internal temperatures. Grate
which satisfies this condition is determined by an iterative heat losses are determined from specified strand inlet
procedure. Calculations are completed for each zone in turn, temperatures and those calculated for the end of the pre-
starting from the discharge end of the kiln. heat fumace.
Table II. Leakage, and Burner Primary Air Rates Used in All Simulations
kg/min.
Drying furnace: hood 30
windbox 150
Preheat furnace: hood 130
windbox 200
Preheat fan bleed-in dampers in closed position 80
Heat recup, bleed-in damper in closed position 30
Kiln: feed end 50
discharge end (includes firing hood) 200
Kiln burner primary air 170
J
Figure 2, the rates of diffusion of oxygen (species A) [A1.10]
through the gas boundary layer and the outer layer of Fe203
(species B) of a pellet, respectively, are given by:
where
( ~ 0 ~ )B-s = 4 7r ~ k,,(YoB- Yos) [AI.1] Al = pl/(bl km ~) and Az = pz/(b2km ~)
()),, 4 [A1.2]
--Bl'l dO + B2 12 dO ~ - = (Yos - rOl)
As for Eq. [A1.4], the rate of diffusion through the second D~2 = D~I, gives:
product layer (hematite and carbon) is:
r2 rl
Oxygen concentrations at the surface (Yos) and solid
The amounts of oxygen which have passed through the interfaces (Yol and Yo2) can be eliminated by combining
layers are directly related to the changes of the interface Eqs. [AI.ll] and [A1.9], and Eqs. [A1.12] and [A1.8],
radii, i.e., respectively, to give: