Upload
ledien
View
214
Download
0
Embed Size (px)
Citation preview
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 163
INCREASED PRODUCTIVITY OF ZINC ROASTERS AND SO2 -
QUALITY
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen
[email protected], [email protected],
Outotec Research Oy, PO Box 69 Pori, FI-28101 Finland
Sasu Penttinen
Boliden Kokkola Oy, PO Box 26 Kokkola, FI-67101 Finland
Kurt Svens
Outotec Oyj, PO Box 86 Espoo, FI-02200 Finland
Bernd Kerstiens
Outotec GmbH, D-61440 Oberursel, Germany
Abstract
Advanced optimizing control of the fluidised bed zinc roaster has been in use for an
extended period at Boliden Kokkola and the implications are seen as extended bed
stability, increased throughput and far better on-line availability. The roaster furnace
control system developed is based on a dynamic and predictive furnace model, with
parameters determined experimentally at the site. The controller maximises the
concentrate feed and keeps the furnace operational parameters inside limits given by
raw material mix and equipment constraints.
The products of a zinc roaster, calcine, sulphur dioxide and the side products, such as
mercury, meet gradually more stringent requirements, in particular as to their assays
and total recoveries in the process chain. The end users of sulphuric acid in northern
Europe, among others in the pulp and paper industry, require mercury levels much less
than the conventional 0.2-0.4 ppm Hg. Therefore, secondary de-mercurising circuits
have been used over the last five years for further cleaning the SO2 stream allowing
production acid grades with less than 0.08 ppm Hg.
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 164
1 Introduction
Boliden Kokkola Oy is a custom smelter and it uses 10-15 different concentrates each
year and a number of other raw materials as components of the feed mixture. The
today’s zinc production uses two main technologies at the site. The roasting-leaching-
electrowinning route has been in use since late 1960s and the more recent atmospheric
direct leaching [1] as the first treatment step to the sulphidic concentrates, has been in
use since 1998 and comprising today about 40% of total zinc production. Two parallel
technologies give clear advantages and increased flexibility in the use and selection of
raw materials. The produced sulphur dioxide gas is led from the roaster to sulphuric
acid plant by Kemira Oy.
Steam drum
Roaster furnace
Boiler
Cyclones Electric
precipitators
ST LH
Ball mill
Calcine
PT HT
Cooling
Air
Water
Steam
KT
VIT LIT
Stack
Heat exchangers
Contact tower
Air in
Air out
Cooling water
NaHS
Weak acid basins
Funda
Lime
To sea
Buffer
tank
WaterHeat recovery
Water
H2O
H2O
98.8% H2SO4
H2SO4
98.8%
95.6%
95.6%
OUTOKUMPU ZINC ROASTER AND KEMIRA SULPHURIC ACID PLANT
Zn-concentrate
Air
ORC / HVT 2/2000
Weak acid tank
H2O
H2OWater lock
ST = Sulphatating Tower
LH = Weak Acid Tower
SiO2
HCl
PT = Washing tower HT = Halogen tower
Se solution
KT = Drying tower
VIT = Intermediate
absorber
LIT =Final absorber
Figure 1. A schematic flow sheet of the combined zinc roaster-acid plant in Kokkola,
by Boliden Kokkola Oy and Kemira Oy, respectively.
Over the recent years, an extensive development program has been carried out for
increasing the treatment throughput of the two 72 m2 fluidized bed (FB) lines. The
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 165
production equipment is basically the same as in 1969 when the smelter was started up
[2], Figure 1, but now significant improvements have been gained in the on-line
availability and production rate.
The development program included laboratory-scale and industrial-scale test runs on
the roasting kinetics and the mechanisms of selected sulphidic zinc concentrates, the
factors behind the bed instability phenomena, as well as introduction of a new advanced
control system to the two FB roaster lines. Improvements were also made in the roaster
gas handling, where the emphasis has been in reducing the mercury traces in the
product acid to a new level. This has been accomplished by the primary scrubbing
stages at the roaster were optimized and a secondary scrubbing step was added to meet
the strict requirements of pulp and paper industry.
2 Roaster gas handling
2.1 The gas train
The roaster gas train post the FB furnace consists of a waste heat boiler, cyclones and
two ESPs before the de-mercurising steps. The gas duct from the roaster to the acid
plant is long, about 1.0 km, separating thus the first scrubbing stage within the roaster
area and the second scrubbing stage at the acid plant, see Figure 1.
2.2 De-mercurising steps
The gas train consists of two separate de-mercurising steps, and allows thus production
of high-purity sulphuric acid from all zinc concentrates, sulphidic and oxidic,
independently of their mercury content. The first step is carried out in the immediate
vicinity of the roaster, and the second step is an additional and new feature of the
washing section of the sulphuric acid plant, see Figure 1.
The first de-mercurising step is composed of two physical stages. In the first stage the
de-dusted roaster off-gas post ESP is cooled by direct water quenching to below 200 °C
and then in the second stage it is scrubbed with a carefully controlled sulphuric acid
solution containing chloride and selenium [3] at about 60 °C. The mercury scrubbing
chemistry is based on the reactions:
Hg°(g)+Hg2+
(aq)=2Hg+(aq) (1)
SeO2
2-(aq)+SO2(g)=Se°+SO4
2-(aq) (1’)
Hg°(aq)+Se°=HgSe (1’’)
2HgSe+Hg+(aq)+2Cl
-(aq)=Hg3Se2Cl2 (1’’’)
In a strongly complexed aqueous medium, containing a spectrum of chloride-complexes
of mercury, as shown in Figure 2, the de-mercurising reactions must thus be capable to
remove both elementary and oxidized forms of mercury from the gas. In spite of the
fact that the stable form of mercury in the roaster off-gas is an oxidized substance,
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 166
elementary mercury is typically determined in such circumstances and it forms a
significant fraction of total mercury [4].
A comprehensive thermodynamic model was built for the scrubber liquid using the
Pitzer model [5], including the complexation of mercury with chloride ions in a strong
sulphuric acid matrix. Description of the acid matrix was taken from a study where
water-sulphuric acid has been modelled to high concentrations and temperatures [6].
The gas phase was assumed to be ideal. The thermodynamic data for the double salt,
equation (1’’’), are not available and thus its saturation limit in the various operational
points could not be optimized.
A typical acid strength of 33-36% H2SO4 is used in the counter-current scrubbing at the
roaster. The other operating parameters of the step have been optimized by extensive
thermodynamic modeling of the gas-liquid –system and by observations at the site on
the mercury content of the off-gas and its speciation between metallic mercury vapour
and nonmetallic gaseous compounds. The formed mercury-selenium double salt is
separated from the scrubber liquid by filtering. The first scrubbing stage is also used for
controlling halogen carry over to the acid making, as the vapor pressured of HCl(g) and
HF(g) are very sensitive to the acid strength and the prevailing outlet temperature of the
scrubber.
1.0E-21
1.0E-19
1.0E-17
1.0E-15
1.0E-13
1.0E-11
1.0E-09
1.0E-07
1.0E-05
1.0E-03
1.0E-01
1.0E+01
100 150 200 250 300
Temperature ( °C )
Mola
lity o
f specie
s (
mol/kg )
Hg(+2a)
HgCl(+a)
HgCl2(a)
HgCl3(-a)
HgCl4(-2a)
Hg(a)
HgOH(+a)
Hg(OH)2(a)
Figure 2. Speciation of aqueous mercury in the scrubber liquid, at the cooling stage in
the first gas purification step at Kokkola zinc roaster, calculated by the thermodynamic
model; the thermodynamic data were taken from [6,9].
The second de-mercurising step is carried out using carefully controlled chemistry and
proper ‘side-reactions’ in the washing section of the acid plant [7]. Instead of pure
water, slightly acidified aqueous solution containing chlorides and selenium is used.
The chemistry is basically the same as shown in equations (1), but the actual process
conditions are different, in particular the scrubber temperatures.
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 167
This solution chemistry has been successfully used at Kokkola since 2001, and the
purity of product acid is typically 0.05 ppm Hg (mg/kg H2SO4), which is much lower
than obtained with the calomel process [8]. In addition to the actual mercury
concentration of each feed mixture, the selenium-to-mercury –ratio is currently actively
controlled at Kokkola, and displayed in the daily roasting reports and instructions.
3 Development work for the operational window of the roaster
In order to improve the stability of the roasting process and to increase the on-line
availability and thus the throughput of the roaster, an extensive series of development
projects was carried out in 1998-2006. It included several limited tests series on the
industrial scale in the roaster area [10-11], dealing with the dynamics of the FB furnace
as well as with the behaviour of various feed mixtures in the roaster line. Various raw
materials were also studied on the laboratory scale [12], in order to get detailed
information about the accretion formation in the roaster vessel and the industrial
roasting kinetics. The overall procedure used is depicted below as the flow chart [13].
3.1 De-mercurising steps
The FB roasters at Kokkola use bed overflow only for the calcine outlet from the
furnace, in addition to WHB and cyclone products. Thus the physical properties of the
furnace calcine produced must meet the requirements of smooth material flow from the
overflow gates and no oversize agglomerates are allowed.
All laboratory scale investigation on the FB roasting of zinc concentrates were carried
out using a small batch-wise operating fluidized bed furnace with an internal diameter
of 40 mm and a typical batch size of 100 g. As the feed mixture in the roaster is
agglomerated by itself or assisted by water additions to conveyer belts or storage bins,
most high-impurity concentrates were additionally roasted as agglomerates in a
horizontal tube furnace [11]. The reason for it was to find out the details of the roasting
mechanisms and origin for the growth of agglomerates during FB roasting.
Mineralogical Analysing Laboratory Plant
evaluations calcine experiments experi-
samples ments
Applications and
operating rules
Thermodynamic Heat balance Active surveillance
evaluations calculations of special situations
in the Roaster
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 168
0
2
4
6
8
1 0
1 2
1 4
1 6
4 9 5 0 5 1 5 2 5 3 5 4 5 5 5 6 5 7 5 8 5 9 6 0 6 1 6 2 6 3 6 4 6 5 6 6 6 7 6 8 6 9
Fe w
t%
Zn w t%
S a m p le
A P irs a
Ag u a s T e n id a s
Azn a c o l la r
B ru n s w ic k
C a ye l i
C o lq u ir i
F e i ta is
G a lm o y
G a rp e n b e rg
H -q ty
H e l lye r
L a N e g ra
L is h e e n
M o in h o
N -M iX
N -M ix
P e ru M ix
P ic k L a k e
P o la r is
R e d D o g
S o tie l
T a ra
T rze b io n k a
Figure 3. The analysed iron concentrations in the sphalerite lattice of selected sulphidic
zinc concentrates determined by EPMA.
Figure 4. Copper-rich core (light) formed in a partially roasted zinc concentrate
agglomerate as a result of rapid migration of impurities in a Cu-rich zinc sulphide.
The results obtained gave a lot new information about the behaviour of several
impurities in the oxidizing sphalerite concentrate, as to the bed non-stability phenomena
and furnace build-up formation [2,10-11]. The micrometer-scale phenomena observed
within the primary ZnS grains and on their oxidising surfaces at roasting temperatures
indicate, for example, that some common components of the zinc concentrates cannot
be the origin of bed non-stability, as typically iron dissolved in the sphalerite lattice,
Figure 3, which also indicates a specific type of sulphide mineralogy in the concentrate
where iron sulphides are of pyrrhotite type instead of pyrites. The detailed process
mineralogical structure analyses of partially roasted sulphides indicated the prevailing
oxidation mechanisms that involve copper and lead, if present at high concentrations,
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 169
migrating in the system. Due to their low melting point compounds they are prone to
forming sinters and thus agglomerating the calcine, Figure 4.
3.2 Impurity control in the feed mixture – heat generation and oxygen
requirement
As various impurities in the concentrates behave in a different manner in the FB
roasting process, their active control is necessary [13]. That typically involves
controlled and limited concentrations for some key elements, as well as for their sum
concentrations and restricted operational temperatures, based on the feed mixture assay.
Therefore, a number of components in the feed mixture have their specific upper
(maximum) limits allowed, as generally used in this industry [14], but also the
mineralogical features of some elements in the concentrates are taken into account
when selecting the feed mixture assay. This typically refers to the distribution of such
elements and their presence in stable, high-melting compounds and less stable phases,
melting around the roasting temperature.
0
200
400
600
800
1000
1200
1400
1600
1 3 5 7 9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
Concentrate source#
Heat value kWh/dmt
O2 requirement Nm3/dmt
SO2 produced Nm3dm/t
Fe*10 wt%
Figure 5. The heat value and oxygen requirement of selected sulphide concentrate raw
materials used in zinc production at Kokkola, organized according to their energy
content [16].
An important and general feature of the FB roaster is its essentially fixed cooling
capacity, which is not a function of the production rate. As the volumetric process air
flow rate is essentially determined by the desired oxygen coefficient of the FB (true
oxygen feed rate to stochiometric oxygen need of the fed mixture) the cooling effect of
the process gas can not be used for controlling the furnace temperature. For this
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 170
requirement, secondary air introduced above the FB can be used if high bed
temperatures are allowed, but the off-gas should be cooled and possibly after-burned to
lower the residual sulphidic sulphur level of the boiler calcine [15]. This feature is,
however, dependent on the bed calcine outlet, whether it is realized as bed underflow or
overflow. The former design ends up with a higer sulphide sulphur in the calcine, due to
the mixing properties of the fluidized bed.
For stabilising the feed mixture preparation routines, the generation of heat in the
roasting (the oxidation enthalpy of the minerals in the feed mixture to the typical
substances in the calcine) and the amount of oxygen needed (Nm3/dmt) in the roasting
is estimated at Kokkola for all the concentrates and other raw materials used, see Figure
5. The roasting heat generation and oxygen consumption values are determined from
the chemical assay and a mineralogical characterization done from time to time for each
raw material.
4 Furnace control development
Controlling the FB roasting optimally is demanding, since many of the process
variables interact. Thus there is a need for multivariable control strategies, and
development of such requires good knowledge of the dynamic dependencies between
input and output variables. In 2003-2004, a model for the furnace dynamics was
developed [17-18] based on data from plant experiments at Boliden Kokkola. Based on
the dynamic model, a multivariable control application was developed and implemented
in 2005. The controller maximises production within the limits given by process
metallurgy and the equipment.
4.1 Measurements and control variables
Furnace bed temperatures are measured in various locations at the bottom and at the
wall of the lower part of the furnace. The bed temperature has to be kept within certain
ranges to maintain proper conditions for roasting. The temperature at the upper part of
the furnace, which is measured at the boiler inlet, is also important. The concentrate
feed is the primary manipulated variable. Also two water feeds are adjusted: water is
added to the concentrate feed on the load conveyor, but can also be fed directly into the
furnace. Air, which is both the fluidising medium and the source for oxygen, is
normally kept at a fixed value. There is a possibility to feed technical oxygen to the
furnace, although this is option currently not used.
The oxygen coefficient, i.e. the ratio between the actual oxygen feed and the amount of
oxygen needed stoichiometrically to combust the feed mixture, should be kept within
ranges given by the concentrate properties. Current plant practise is to control an
estimate of the oxygen coefficient, based on the oxygen demand of the rated
concentrate mixture. There is a measurement of the oxygen content in the boiler off-
gas, which gives information of the realised oxygen coefficient. However, this
measurement is somewhat uncertain due to air leakages into the furnace and the boiler.
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 171
4.2 Dynamics
A large number of step-change experiments were carried out to identify the dynamic
behaviour of the FB furnaces. The experiments comprised of changes in six
manipulated variables, see Table 1. During the experiment periods, the existing
automatic controllers for furnace temperature and boiler inlet temperature were
switched off to obtain data from the open loop system. The sampling time was chosen
to ten minutes, since the data collection system at the roaster plant did not store values
more frequently.
The observed effects are qualitatively shown in Table 1. The dependencies are highly
interconnected. For instance, not only the concentrate feed but also the feeds of water
and air affect the furnace temperature. An example of the effect of changes in load
conveyor water feed is shown in Figure 6. The load conveyor water has a rather
complex effect on the furnace bed temperature, the long-term effect is positive, but
there also is an inverse response during the first hour.
Table 1. Dependencies between manipulated and observed variables in the fluidised
bed furnace.
Effect on
Step-change
increase in
Bed
temperatures
Boiler inlet
temperature
Offgas oxygen
content
Windbox
pressure
Concentrate feed increase increases decreases -
Air feed decrease increases increases increases
Oxygen feed diverge - increases increases
Overflow grate area
pressure
- - - decreases
Load conveyor
water feed
temporary
decrease, then
increase
decreases temporary
increases
-
Furnace water feed decrease - - -
Quantitative modelling of the dynamic effects shown in Table 1 was made using
ARMAX type time series models. It was known that this linear model structure could
not describe the furnace behaviour in all situations, e.g. when the oxygen coefficient is
shifted to values below one. However, the primary goal was to develop a model for
normal operation conditions, where a linear description was assumed to be adequate. As
a first step in the identification, partial models (sub-models) were identified for the
single input-output pairs. The aim was to find models as simple as possible describing
the dependencies between input and output observed in the experiments. Zero order and
different first order models were used. Finally, the partial models were combined into
one model for each output, which also included disturbance models.
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 172
0
100
200
300
400
500
600
L/h
Load conveyor water feed
920
930
940
950
960
06:00 09:00 12:00 15:00 18:00
°C
Furnace temp. 6 Furnace temp. 7
Furnace temp. 12 Furnace temp. 14
Figure 6. The effect of step changes in load conveyor water feed on temperatures in the
furnace bed.
In addition to the step-change experiments, tracer tests were carried out to clarify the
residence time of the solid material in the bed and in the gas line. In the tests, TiO2
material with a density and a particle size distribution close to that of calcine was used
as tracer. The tracer was fed in impulses (batches) to the feed mixture at the slinger belt
feeder. The results indicate that the mean residence time of the solid particles in the bed
is approximately 20 hours, as can be seen in Figure 7. For the gas line, however, very
short residence times were observed. The peak concentration in the cyclone (post
boiler) was obtained around one minute after the impulse in the furnace feed.
0
0.05
0.1
0.15
0.2
0.25
-10 0 10 20 30 40 50 60 70 80 90
Time (h)
Ti (%
)
2.9.2003
22.9.2003
model
Figure 7. Tracer content in the bed overflow in two experiments: the tracer was fed as
an impulse at time 0 h. A first order model with a time constant of 20 h is shown for
comparison.
The results are quite similar to the results by Dimitrov’s [19] in a small FB furnace.
Also Themelis [20] has reported results of retention time tests, using a radioactive
tracer. He showed experimentally that various particle size fractions have different
retention times.
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 173
5 Control system
A control system was developed to enable real-time production maximisation within the
limits given by process metallurgy and equipment. The input and output signals of the
control application are shown in Figure 8. The optimising control system maintains a
concentrate feed as high as possible so that all variables are within given limits. Water
additions (to the concentrate on the load conveyor and/or directly to the furnace) are
used if they enable a higher concentrate feed.
User specified limits
Concentrate feed
Load conveyor water feedf
Furnace water feed
Bed temperature
Boiler inlet temperature
Temperature of air feed
Oxygen feed
Oxygen requirement of concentrate
Concentrate feed
Load conveyor water feed
Furnace water feed
Air feed
control
model
Oxygen coefficient (min) Concentrate feed (max) Load conveyor water (min, max) Furnace water (min, max) Bed temperature (min, max) Upper temperature (min, max) Offgas O2 (min)
Figure 8. Input and output signals of the optimising control system.
As described in previous chapter, the interactions in the fluidised bed furnace are cross-
coupled. Thus, a MIMO (multiple input – multiple output) control structure was used,
since separate single-input, single-output controllers would not operate optimally. The
control system includes a state estimator based on the dynamic ARMAX models
identified in the step-change experiments, an optimisation routine based on the
SIMPLEX algorithm, and a controller based on state prediction. The algorithms were
developed in MATLAB®
, which is also the environment used for control system
implementation. OPC is used for the communication between the control system on a
PC and the Honeywell TP®
Alcont automation system of the roaster. The operator
interface, where the limits are specified, is in Alcont.
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 174
M O Q S U NM NO NQ NS
OO
OQ
OS
OUÅçåÅÉåíê~íÉ=ÑÉÉÇI=íLÜ
M O Q S U NM NO NQ NS
PKU
Q
QKO
ñ=NMQ ~áê=ÑÉÉÇI=kãPLÜ
M O Q S U NM NO NQ NSM
RMM
NMMM
Ñìêå~ÅÉ=ï~íÉêI=äLÜ
M O Q S U NM NO NQ NSVMM
VOM
VQM
ÄÉÇ=íÉãéÉê~íìêÉI=…`
M O Q S U NM NO NQ NSMKV
N
NKN
çñóÖÉå=ÅçÉÑÑK
a~óë
éêçÅÉëë
äáãáí
N O
Figure 9. Data from the model-based control of a FB furnace. See text for explanation
of periods 1 and 2.
The calculation interval of the control system is 5 minutes, and the main steps in a
calculation round are:
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 175
1. Pre-treatment and validity check of process data
2. Updating the state of the dynamic model
3. Searching optimal set of manipulated variables
4. Control actions based on 30 minutes prediction.
Figure 9 shows history trends from 20 days operation of a furnace, and the actions of
the control system. In period 1, the upper limit for bed temperature limits the
production, and the feed of furnace water is at maximum limit. In period 2, the lower
limit for oxygen coefficient limits the production.
6 Conclusions
The complex nature of material flows and the increased requirements as to their quality
and on-line availability, in particular in a large industrial network of several
independent companies, are a major challenge for the metallurgical process industry.
Therefore, an intensive R&D programme has been accomplished at Boliden Kokkola
zinc smelter for securing the quality issues in the zinc production, including also its by
products such as sulphuric acid.
References:
1. Takala H, Leaching of zinc concentrates at Outokumpu Kokkola plant, Erzmetall, 52 (1999) 1, 37-42
2. Metsärinta M-L, Taskinen P, Nyberg J, Ohvo E, Industrial Scale Roasting of Impure Concentrates at Boliden Kokkola, Proceedings of PbZn 2005 – International Symposium of Lead & Zinc Processing (Ed. T Fujisawa, J Dutrizac, A Fuwa, N Piret, A Siegmund), MMIJ, Kyoto (2005) pp. 343-357
3. a. Bryk P, Haani M, Honkasalo J, Kinnunen J, Lindsjö O, Nyholm E, Poijärvi J, Rastas J, Kangas J, Process for removal and recovery of mercury from gases, USP 3 677 696, Outokumpu Oy, appl. July 3, 1970, and b. Kuivala AMF, Leppinen JO, Poijärvi JK, Rastas JK, Process and apparatus for the removal of mercury from sulphur dioxide-bearing hot and moist gases, USP 4 579 726, Appl. Jul. 9, 1984
4. Nishitani T Fukunaga I, Itoh H, Nomura T, The relationship between HCl and mercury speciation in flue gas from municipal solid waste incinerators, Chemosphere, 39 (1999) 1, 1-9
5. Pitzer K, Thermodynamics of Electrolytes. I: Theoretical basis and general equations, The J. Phys. Chem., 77 (1973) 2, 268-277
6. H Sippola, Modelling the solubility of ferrous sulphate in sulphuric acid-water solutions, Helsinki University of Technology, Licentiate thesis, 1992, 109 p.
7. Peltola H, Taskinen P, Takala H, Nyberg J, Natunen H, Panula J, Method for removing mercury from gas, WO 02/45825 A1, Outokumpu Oyj, appl. Dec. 8, 2000
8. Hultbom KG, Industrially proven methods for mercury removal from gases, EPD Congress 2003 (Ed. ME Schlesinger), TMS, Warrendale (PA), 2003, pp. 147-156
9. Roine A, HSC Chemistry for Windows, vers. 5.0, Outokumpu Research Oy, Pori, report 02103-ORC-T, 12.08.2002, 363 p.
The Southern African Institute of Mining and Metallurgy Lead and Zinc 2008
Pekka Taskinen, Maija-Leena Metsärinta, Björn Saxen, Sasu Penttinen, Kurt Svens, Bernd Kerstiens
Page 176
10. Taskinen P, Jyrkönen S, Metsärinta M-L, Nyberg J, Rytioja A, The behaviour of iron as impurity in zinc roasting, Sulphide Smelting 2002 (Ed. R Stephens & HY Sohn), TMS, Warrendale (PA), 2002, pp. 393-404
11. Metsärinta M-L, Taskinen P, Jyrkönen S, Nyberg J, Rytioja A, Roasting mechanisms of impure zinc concentrates in fluidized bed, Yazawa International Symp., Vol 1. (Ed. F Kongoli, K Itagaki, C Yamauchi & HY Sohn), TMS, Warrendale (PA), 2003, pp. 633-646
12. Metsärinta M-L, Taskinen P, Nyberg J, Ohvo E, Mechanisms of impure zinc concentrates roasting in fluidized beds, Tsvetnye Metally, 6 (2005) 5-6, 92-99
13. Nyberg J, Characterisation and control of the zinc roasting process, PhD Thesis, University of Oulu, 2004, 114 p.
14. Grant RM, Zinc concentrate and processing trends, Int. Symp. – World Zinc ’93, (Ed. IG Matthew) Aust. IMM, Victoria (Aust.), 1993, pp. 391-397
15. Svens K, Kerstiens B, Runkel M, Recent experiences with modern zinc processing technology, Erzmetall, 56 (2003) 2, 94-103
16. Metsärinta M-L, Thesis for Lic. Techn. degree, University of Oulu, 2008 17. Härkönen, J. Identification of a Zinc Roaster Furnace, Master´s thesis. Åbo Akademi
University, Finland, 2003. 18. Saxén, B., J. Nyberg, J. Härkönen, H. Toivonen, Dynamic Behaviour of a Zinc Roaster
Furnace, 11th IFAC Symposium on Automation in Mining, Mineral and Metal processing - MMM 2004, September 8-10, 2004, Nancy, France
19. Dimitrov R, The improvement of zinc concentrates roasting in a fluidized bed, Rudarsko-Metalurški Zbornik, 30 (1983) 4, 397-419
20. Themelis NJ, Freeman GM, Fluid bed behaviour in zinc roasters, JOM, 36 (1984) Aug., 52-57