Sie sind auf Seite 1von 18

Mineral Processing and Extractive Metallurgy

Transactions of the Institutions of Mining and Metallurgy

ISSN: 2572-6641 (Print) 2572-665X (Online) Journal homepage: http://www.tandfonline.com/loi/ympm21

BOF process dynamics

Ghosh Snigdha, Ballal N. Bharath & Nurni N. Viswanathan

To cite this article: Ghosh Snigdha, Ballal N. Bharath & Nurni N. Viswanathan (2019) BOF
process dynamics, Mineral Processing and Extractive Metallurgy, 128:1-2, 17-33, DOI:
10.1080/25726641.2018.1544331

To link to this article: https://doi.org/10.1080/25726641.2018.1544331

Published online: 27 Nov 2018.

Submit your article to this journal

Article views: 35

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=ympm21
MINERAL PROCESSING AND EXTRACTIVE METALLURGY
2019, VOL. 128, NOS. 1–2, 17–33
https://doi.org/10.1080/25726641.2018.1544331

FERROUS CONVERTERS

BOF process dynamics


Ghosh Snigdha, Ballal N. Bharath and Nurni N. Viswanathan
Centre of Excellence in Steel Technology, Department of Metallurgical Engineering and Materials Science, Indian Institute of Technology
Bombay, Mumbai, India

ABSTRACT ARTICLE HISTORY


The Basic Oxygen Furnace (BOF) process is an exceedingly fast refining process needing a good Received 28 July 2018
dynamic model for better understanding and dynamic control. The process is characterised by Revised 31 October 2018
reactions at multiple scales: at the scale of the metal bath and the slag and at the scales of Accepted 1 November 2018
droplets and bubbles. The reactions take place at multiple sites too. The presence of the
KEYWORDS
supersonic jet interacting with the metal bath and the slag layer, producing various sizes of Basic Oxygen Furnace; BOF;
drops in the emulsion which on reaction produce copious bubbles at its interface, lime process; dynamics; emulsion
dissolution issues, etc., make the description of the dynamics of the process complex and mechanism; decarburisation;
fascinating. There are still areas where an adequate quantitative description needs further dephosphurization
research.

Introduction and bubbles distributed in the slag/metal/gas emulsion


phase. The difference in length scales result in differ-
In integrated steel plants, Basic Oxygen Furnace (BOF) ence in time scales too: the metal bath sees changes
plays a predominant role. Of the total crude steel pro- over the entire heat cycle of 12–15 min whereas the
duction of 1.6 billion tonnes in 2016, almost 75% was drops may undergo the complete cycle of refining in
made through oxygen blown converters, most of about a minute. Therefore, the picture of the process
which are top blown BOF (World Steel Association dynamics has evolved over many years based on obser-
2017). The productivity, economics and quality of vations and measurements in commercial and pilot
steel produced through oxygen blown converters plants, carefully designed experiments and mathemat-
have a large bearing on the health of the steel industry ical modelling.
worldwide. These aspects are determined by the raw
material (hot metal) quality on one hand and the
dynamics of the process on the other. BOF process is
The process
characterised by high reaction rates, the refining pro-
cess being completed typically in 12–15 min. To con- A typical BOF converter consists of a cylindrical barrel
trol this process for quality and productivity in this with a rounded bottom, and a conical top (25–30° half
short time frame, a good understanding of the cone angle) for directing the gases into the off-gas hood
dynamics is important. (Figure 1). The body is supported on pivots, called the
BOF is a complex process taking place over a short trunnions, such that the furnace can be rotated around
time span, with very little direct feedback information for charging, sampling, tapping and slag-off. The inside
as the process progresses. There are several sub-pro- is typically lined with magnesia-carbon refractory, of
cesses which are either ill-understood or of which we different quality and thickness to match the wear pat-
have only semi-quantitative understanding. It is an tern. The typical volume provided inside the vessel is
autogenous process; there is a heat excess even after around 1 m3 per ton of steel produced. If the slag
the input hot metal at around 1350°C is delivered as weight is 100–120 kg/ton, the freeboard above the
steel at 1650–1700°C. Various coolants are therefore quiescent bath is more than 80%. This accommodates
used, scrap and iron ore being the primary ones. Oxy- the vigorous reactions which take place during the
gen is delivered to the process through supersonic jets middle part of a typical blow. The bottom of the reactor
issuing into hot and dust laden gases or under a liquid is fitted with several (typically 6–8) porous elements,
gas emulsion; the jet behaviour is affected by this ambi- through which argon gas is passed for bath mixing
ent. The reactions are heterogeneous and at different and aiding slag-metal reactions. A tap hole is provided
length scales. There are the bulk metal bath, the bulk on one side in the lower part of the cone for tapping the
slag and the gas phases. Much of the reaction, on the metal. Slag is poured out on the other side through the
other hand, takes place at the scale of fine droplets mouth.

CONTACT Nurni N. Viswanathan vichu@iitb.ac.in Centre of Excellence in Steel Technology, Department of Metallurgical Engineering and Materials
Science, Indian Institute of Technology Bombay, Mumbai, India
© 2018 Institute of Materials, Minerals and Mining and The AusIMM Published by Taylor & Francis on behalf of the Institute and The AusIMM
18 G. SNIGDHA ET AL.

top lance fitted with 3–6 supersonic flow nozzles


(2.0–2.1 Mach, fitted at an angle to the lance axis).
The tip of the lance is held at a distance between
1.8–2.5 m above the level of the quiescent metal
bath in a large sized converter. This lance height is
one of the operating parameters to control the
process.
A typical tap-to-tap cycle consists of the following
steps. Charging sequence is: lime, scrap and hot metal.
Once the furnace is made upright, the oxygen lance is
lowered to the required height (initially, highest value,
2.2–2.5 m) and blowing is started. During the initial
half of the blow, additional lime if any along with
iron ore, dolomite, and any other additives are
added. Additions of moisture containing materials
are avoided during the latter part of the blow to
keep hydrogen in the steel produced low. The high
lance operation is continued (typically 3–4 min) till
the slag has enough FeO for facilitating lime dissol-
ution, as discussed later. Thereafter the lance is pro-
gressively lowered to achieve the required rates of
refining. The lance height is decreased in 3–5 steps
depending on the individual plant practice. At
around 80–90% of the blow (based on oxygen flow)
a sample is taken for analysis and temperature is
measured so that when one finishes the blow
required composition and temperature are achieved
simultaneously. Sampling may be done either manu-
ally, i.e. stopping the blow, turning the converter to a
near horizontal position and taking a sample through
Figure 1. Schematic of BOF. a spoon and measuring the temperature, or through
a sub-lance which is lowered into a blowing furnace
The primary raw material is molten pig iron (hot (in-blow sampling). Based on the sample analysis
metal) at about 1300–1400°C. Since heat generated is and temperature, the last part of the blow is com-
more than what is required, steel scrap along with pleted with requisite trim additions. Once the blow
iron ore is used as coolant. Lime stone is added in is completed the converter is turned to the tapping
some shops as a coolant for adjusting the final tempera- side for pouring the metal out, and then to the
ture. Calcined lime is used as a flux for achieving high other side for slag-off. In modern practice, some
basicity needed for phosphorus removal. Scrap is added slag is retained, the furnace is made upright, some
first to an empty vessel (after slag-off from the previous magnesite is added and then the slag is splashed
heat), on which the requisite quantity of hot metal is on to the inside surface by blowing high velocity
added. Iron ore when used is added in a distributed nitrogen. Periodically, the empty furnace is inspected
fashion typically during the first half of the blow. for refractory damage, either manually or through
Part or all of required lime is added before scrap laser scanners. Damage is repaired by gunning
addition to act as a pad for protecting the lining from refractory powder. The furnace is then ready for
scrap fall. Rest of the lime is typically added in a distrib- the next blow (Miller et al. 1998).
uted fashion during the blow. Some magnesia addition In modern steel shops with level 2 automation,
takes place in the form of dolomite, to minimise refrac- much of the blow control is automatic. Normally
tory dissolution into the slag. The amounts of various there are several blow profiles which are stored.
charge materials are calculated a priori by a charge Depending on the input/output conditions, the model
control model based on material and heat balance, suggests the blow profile, when an operator can utilise
taking into account the input compositions, hot her discretion and select. Thereafter the controller
metal temperature and output steel composition and completes the blow making appropriate additions
temperature. and lance height variations. The operator still needs
The refining reactions are all oxidising. This is to supervise and intervene when needed, since the pro-
accomplished by blowing tonnage oxygen through a cess has some stochasticity.
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 19

Reactions
Typical composition of the hot metal can be: 4.5% C,
0.3–0.5% Si, 0.2–0.7% Mn, 0.1–0.18% P, 0.02% S, and
temperature of 1350°C. Since sulphur can only be
removed to the slag in the reduced state in the presence
of liquid iron, the oxidation process in the BOF does
not remove any significant amount of sulphur. The
overall reactions of significance can be written as fol-
lows:
[Si] + {O2 } = (SiO2 )

1
[Mn] + {O2 } = (MnO)
2
1
[C] + {O2 } = {CO}
2
5
2[P] + {O2 } = (P2 O5 )
2
1
Fe (l) + O2 (g) = (FeO)
2
[],{} and () are used for the metalloids dissolved in the
metal bath, gas and constituents in the slag, respect-
ively. Figure 2 gives the progress of the reactions in a
200 T converter as measured by Ciccuti et al. (Cicutti
et al. 2000, 2002). Earlier measurements too show simi-
lar above patterns (Chatterjee et al. 1976; Miller et al.
1998). Figure 2(c) gives the corresponding evolution
of the slag composition (Cicutti et al. 2000).
One notable feature in the evolution of metal com-
position is the simultaneous removal of significant
amounts of carbon even before Si has dropped to a
very low level. This is also borne out by the observation
that the CO flame shoots up at the mouth of the con-
verter within a short time of start of oxygen blow. This
is in contrast to the observations in the now obsolete
Bessemer converter or the OBM process, where air/
oxygen is blown from the bottom. In the latter pro-
cesses, appearance of a significant flame takes some Figure 2. (a) Evolution of carbon content in the bath during
time, which is assumed to indicate that carbon oxi- the blow. (b) Evolution of other solute contents (c) Evolution
dation does not start till silicon has fallen to fairly of slag composition (Reproduced from Cicutti et al. 2000).
low values (Orban et al. 1961).
Thermodynamically, the order of the solute oxi-
dation reactions at any local location for the input con- analysis of the process dynamics and reaction mechan-
ditions mentioned above should be: Si, Mn, C and isms interesting.
P. That is, at the conditions prevailing in the initial
part of the blow, Si should be oxidised before carbon.
Characteristics of the process
The low initial temperature itself makes Si reaction
to be favourable. In addition the product SiO2 is at Since feedback information from the process is scant,
very low activity in the highly basic conditions main- one needs to build a model for the process dynamics
tained right from the beginning. Carbon monoxide from the characteristics one observes from the infor-
partial pressure on the other hand remains near one mation one can obtain. The following are the typical
atmosphere. For example, if one assumes an activity features:
0.001 for SiO2, the pO2 in equilibrium with 4.5% C
and 0.5% Si is 10−17 and 10−19 atm. respectively. (1) The reaction rates are extremely fast. During peak
These and other features discussed below makes the decarburisation carbon is removed at the rate of
20 G. SNIGDHA ET AL.

about 0.3% per minute; that is about 600 kg of car-


bon per minute in 200 T converter (Figure 2(a)).
(2) The carbon reaction shows three typical phases
(Figure 2(a)): an initial phase when the rate is
building up, an intermediate phase when the rate
is relatively constant in spite of the fact that the
carbon content in the bath continuously falls
from about 3.5–4.0% during this period, and a
final phase beyond a critical carbon content
when the rate tapers off. The critical carbon con-
tent that has been mentioned in the literature lies
between 0.2–0.5%. The value reported by Ciccutti
et al, for example, is about 0.35% (Cicutti et al. Figure 4. Effect of decarburisation rate with oxygen blowing
2000). rate and lance height (Meyer et al. 1964).
(3) Though the three phases are displayed in almost all
studies, the individual heats with identical blowing
conditions show wide irreproducibility (Li et al. (5) Chatterjee et al. (Chatterjee et al. 1976) performed
1964; Meyer et al. 1964) as shown in Figure 3. experiments in a pilot converter at MEFOS
Two sequential blows with identical inputs and wherein they showed that there is a concentration
process parameters can show quite different beha- variation along the height of a top-only blown con-
viours, with some blows displaying slopping verter (Figure 5). This indicates that top blowing
(metal-slag-gas emulsion boiling over the mouth does not mix the metal bath well in spite of the
of the converter) or dry slag and spitting (resulting enormous momentum in the top jet. This differ-
in lance and mouth build-up). The irreproducibil- ence however disappeared on blowing a very
ity was much more prevalent in the early days of small amount of inert gas from the bottom.
the BOF, when bottom blowing of stirring gas (6) It is well known that the slag in a BOF contains a
was not yet incorporated. Use of less scrap as a significant fraction of the metal in the form of dro-
coolant might have also led to more reproducibil- plets in the slag phase. The amounts vary during
ity and decreased slopping. the blow, being highest during the middle part of
(4) After a study of several BOF plants, Meyer et al. the blow (Figure 6, Jalkanen and Holappa 2014).
(Meyer et al. 1964, 1968) showed that the rates The estimates vary between 10–25% (Meyer et al.
of peak decarburisation is directly proportional 1968; Kozakovitch 1969; Chatterjee et al. 1976;
to the rate of oxygen blowing (Figure 4). Simi- Cicutti et al. 2000, 2002; Lee and Kolbeinsen
larly, Kattenbelt and Roffel (Kattenbelt and 2007). These droplets are quite fine, most being
Roffel 2008) showed during the blow in a labora- less than 1–2 mm (Lee and Kolbeinsen 2007;
tory sized converter, that the effects of increasing
rate of oxygen blowing rate and decreasing the
lance height on the rate of peak decarburisation
were similar.

Figure 5. Difference in carbon concentration in top and bot-


Figure 3. Decarburisation rate vs progress of blow for two tom of a 8 T pilot converter during the blow (Replotted
heats (a) (Redrawn from Meyer et al. 1964). using data read from Chatterjee et al. 1976).
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 21

Oxygen jet being almost pure, molecules should


reach the bath surface directly without a significant
mass transfer barrier. When a molecule strikes, it can
do one of the following things:

(a) React with C at the impact site. The reactions


would be:
1
[C] + {O2 } = {CO}
2

Figure 6. Fraction of metal droplets in the emulsion (MFD) and [C] + {O2 } = {CO2 }
foam height (FH) during the blow (Jalkanen and Holappa
2014). (b) It can dissolve in the metal as [O]. This can then
travel elsewhere and react with other oxidisable
elements: O2 = 2[O].
(c) Some of it may react with Fe in the bath producing
Dogan et al. 2011a; Jalkanen and Holappa 2014).
(FeO).
The number of droplets in the emulsion falls
towards the end of the blow. The droplets are 1
Fe + {O2 } = (FeO)
usually in a much more advanced state of refining 2
as compared to the bulk metal bath (Meyer et al.
1968; Cicutti et al. 2000).
This can travel to the slag phase and react with metal
(7) There exists a slag-metal-gas emulsion during
elsewhere.
most of the blow. By about one third of the blow
Each one of these will lead to refining reactions to
the emulsion height exceeds about 2 m, thus sub-
take place at different possible sites in the converter,
merging the lance tip and muting the sound of
leading to, on mixing, overall bath refinement. These
the supersonic jet. Sometimes, the emulsion can
sites are shown schematically in Figure 7.
fill the whole furnace, boiling over the mouth
One should remember that C-O reaction is hetero-
(slopping). Towards the end of the blow beyond
geneous; there is at least one mass transfer step
a critical carbon in the bath, the emulsion collapses
which can be rate limiting. Carbon has to diffuse in
indicating that the emulsion is transitory needing
the metal to the interface. Transfer of oxygen in the
continuous gas generation for its survival (Kozako-
gas phase, dissolved oxygen in the metal phase or
vitch 1969).
FeO in the slag phase may also be involved depending
(8) As mentioned earlier, C, Mn and P are oxidised
on the oxygen source for the reaction.
simultaneously with Si in the initial part of the
Oxygen dissolved may travel to other parts inside
blow, against the expectation of Si reaction being
the metal bath and react with dissolved carbon to
preferred over the other reactions based on bulk
release CO to gas-filled pores in the refractory (site
metal bath composition. Manganese and P reac-
1). CO may also form on solid particles floating in
tions can be explained to some extent by the fact
the metal bath by heterogeneous nucleation (site 2).
of activities in the slag. Carbon reaction cannot
Heterogeneous nucleation may also take place at the
be explained, unless one hypothesises that the
slag layer/metal bath interface (site 3). Homogeneous
bulk metal composition does not prevail at the
nucleation within the bath is highly improbable unless
site of the reaction.
the CO super-saturation is very high (Roddis 1973). As
(9) There is reversal of Mn and P during the middle
mentioned earlier CO reaction may directly take place
part of the blow (Figure 2(b)). This is reflected in
at the impact site (site 5).
the slag path too (Figure 2(c)). It is clear however,
Some of the FeO formed at or near the impact site
that the reversions are correlated with the FeO
may get under the metal bath surface and travel along
content in the slag. CaO dissolution continues
the slag/metal interface, reacting with carbon giving
almost till the end in spite of it being added in
emulsified interface (as in site 3). Bulk of the FeO
the beginning or in the early part of the blow.
formed however probably transfers to the slag
phase. This now gives several possibilities. At the
interface between slag and metal bath, reactions can
Sequence of reactions and reaction sites
take place as mentioned earlier (site 3), oxygen now
Since C determines the overall dynamics of the process coming from the slag phase and carbon from the
as we will see later, we discuss the dynamics with refer- metal.
ence to this reaction taking place vigorously. Other As presented earlier, the slag phase contains a large
reactions can be understood on this framework. number of metal droplets, being continuously
22 G. SNIGDHA ET AL.

Figure 7. Reaction sites in the BOF.

generated by the momentum of the jet at the impact disappears with as little as 1% of inert gas being
site. The FeO in the slag therefore can react with blown from the bottom in comparison to the top gas
these droplets through various mechanisms: flow (Chatterjee et al. 1976), the mechanisms at sites
1 and 2 may be discounted as being unimportant.
i CO bubbles may heterogeneously nucleate at the The temperature at the surface at the impact zone
interface (site 8). is expected to be above 2400 K (Lee and Kolbeinsen
ii CO may be transferred to a passing bubble which 2007). The rates of the chemical reactions are there-
comes in contact with the drop (site 4). fore expected to be very high. The area of the impact
iii CO bubble may homogeneously nucleate inside site is comparatively small and oxygen arrival rate is
the droplet, if the super saturation is very high very high. However the solutes need to diffuse to the
(site 6). interface and the heat has to conduct into the metal.
Fresh metal is brought to the interface which is swept
If some metal droplet is thrown to the free board, it outward by large surface velocity. Under these cir-
may react directly with any oxygen or CO2 in the gas cumstances, one may expect the impact surface
(site 7). being starved of the solutes leaving behind a layer
Though all these sites may be, to some degree, active of Fe reacting with oxygen. Ultimately it may be
during the blow, we need to identify the predominant reasonable to assume that a layer of metal of the
mechanism which determines the overall dynamics. bulk metal composition is entirely oxidised, con-
We can evaluate contribution from each of these sites densed phase oxides being transferred to the slag
based on the observations presented in Section 4. layer. When the carbon content is about 5 wt% (20–
The fact that the bath shows a concentration gradi- 25 mol%), this approximation would mean that
ent in the absence of bottom gas injection, which about 25% of the supplied oxygen is consumed for
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 23

carbon oxidation (CO and CO2) at this site. Dogan gas. If the ambient is lighter the expansion is less due
et al. (2011b) estimated the contribution to be to the mass effect (Alam et al. 2010).
about 40% based on calculations assuming metal In supersonic jets compressibility factor affects jet
side mass transfer not to be rate controlling. At one expansion. Anderson and Johns (1955) have shown
time this was considered to be the major mechanism that the jet doesn’t expand much until the axial velocity
(hot zone or impact zone theory). Metal layer flowing decelerates to the sonic velocity (supersonic core).
outward at this site may also get saturated with oxy- Thereafter the jet expands as a subsonic jet as shown
gen as mentioned earlier. in Figure 8 (Szekely and Themelis 1971). Recent CFD
That leaves us with reactions in the emulsion, which studies of an oxygen jet into a BOF is shown in Figure 9
seem to contain major sites for reactions (sites 4, 6 and illustrating some of these (Odenthal et al. 2006). One
8). The droplets in the emulsion have extremely large can see that the axial velocity is almost constant for a
specific surface area. In the presence of reasonable distance of almost 1 m and the temperature of the
amounts of FeO in the slag, all the refining reactions gas at the axis remains at around −170° C in this
in a droplet can take place in a matter of tens of region. Thereafter the temperature increases steadily
seconds, instead of minutes as seen in Figure 2. A due to entrainment of hot gases. Oxygen lances are
3 mm drop of metal containing 4.5% C, evolves therefore operated at an exit Mach no of about two
about 3000 times its volume in CO. This as it escapes so that it can be kept at some distance and still effect
through the viscous slag emulsifies it. The complex good jet/metal interaction.
interactions of emulsion formation, droplet generation, One may note that if the oxygen jet is submerged in
droplet residence time, etc., thus contribute largely to an atmosphere of CO as in the BOF, the concentration
the overall dynamics. of oxygen can come down substantially.
Reactions of droplets reacting with the gas phase
directly is important primarily in the first couple of
minutes of the blow when a complete slag layer has Gas metal interaction and droplet
not yet formed. generation
A comprehensive look at this overall process
dynamics needs a background on supersonic gas jets, When a high velocity jet hits a metal surface, a crater is
their interaction with a metal/slag bath, droplet gener- formed, the edges of which are highly unstable due to
ation and their residence times, lime dissolution, bath the high velocity of the deflected jet, throwing out
mixing etc., which are discussed briefly below. metal droplets. At high enough values the jet becomes

Gas jets
A gas-in-gas jet entrains the ambient gas in its periph-
ery. The disturbed layer reaches the jet axis a few nozzle
diameters downstream (potential core region) beyond
which the flow becomes fully developed with self-simi-
lar radial velocity profiles. This velocity profile at some
axial distance from the nozzle can be expressed as Figure 8. Supersonic core of an oxygen jet flowing from a con-
(Abramovich et al. 1984) vergent -divergent nozzle.
 
r
−0.694
u ruc /2
=e
uc
where u is the velocity at a radial distance r from the jet
axis, uc is the velocity at the jet axis, and r(uc /2) is the
radial distance from the jet axis where u is uc /2.
The axial velocity varies inversely with distance to
maintain momentum conservation:
k j,u u0
uc =
z/r0
where z is the axial distance from the nozzle. The con-
stant kj,u has a value of about 6.2 (Szekely and Themelis
1971; Abramovich et al. 1984). Typically, the jet Figure 9. CFD simulation of the CD nozzle; Mach number and
expands at a half cone of angle of about 10–12° if the temperature distribution for the design point (Odenthal et al.
ambient gas has the same density as that of the jet 2006).
24 G. SNIGDHA ET AL.

re-entrant, where some of the droplets are thrown into


the jet itself leading to a highly unstable crater oscillat-
ing and rotating around (Molloy 1970). In the presence
of a slag layer these droplets are trapped by the slag
leading to droplet-in-slag emulsion.
The crater depth e can be calculated by performing a
momentum balance at the stagnation point at the cen-
ter of the crater (Bradshaw and Richardson 1970):
k2j,uu2o rg
e= . .
2g (z/ro )2 rl
Turkdogan (1996) modified this with a constant based
on his experiments with various liquids and gases at
room temperature.
Turner and Jahanshahi (1987) studied quantitat-
ively the emulsification phenomena with the help of a
2-D, two phase model of mercury and glycerol. They
found, as expected, that the droplets in the emulsion Figure 10. Effect of top gas flow rate on the droplet generation
increased with gas flow rate and varied inversely with rate (Subagyo et al. 2003).
stand-off distance of the lance tip from the liquid sur-
face (lance height). He and Standish (Standish and He
1989; He and Standish 1990) experimented with a 3-D In a supersonic jet, say of Mach 2, the exit gas tempera-
model of water to determine the droplet generation ture is about 170 K. Thereafter, it entrains lower den-
rate with a top layer representing slag. It was found sity furnace gas. The temperature, velocity and
that there are two regions: one at a lower flow rate composition of the gas change as the jet strikes the
where the rate increases nonlinearly with flow rate bath. Therefore this correlation has large uncertainties,
and the second where the rate varies almost linearly because of which usefulness of the correlation in the
with the flow rate. The Weber number BOF model is less than adequate. Since there is no
other correlation, one normally uses the above corre-
rg u2j lation for generation of droplets and tune it as needed.
NWe = √
sg rl The reaction rates also depend on droplet sizes. Sev-
eral researchers have obtained emulsion samples from
was used to characterise this flow phenomena. Here uj
working BOFs or laboratory hot models. They have
is the nominal gas velocity at the stagnation point in
found in general the sizes to be in the range of 0.05–
the absence of the liquid. Subagyo et al. (2003)
3 mm (Meyer et al. 1968; Cicutti et al. 2000, 2002).
defined a Blow number, NB, to measure the intensity
In the experiments of Koria and Lange (1984) with
of the jet metal interaction-
pig iron and oxygen, there were large chunks of liquids
NB = 0.1 NWe thrown out, which would normally spend negligible
amount of time in the emulsion. Interestingly, they
Droplet generation rate RB (kg/sec) was then corre-
found that the droplet size distribution follows Rosin-
lated experimentally as a function of the blow number
Rammler- Sperling distribution:
as follows-
 n
d
RB NB3.2 R = 100exp(− ′
= (2.11) d
FG 2.6 × 106 + 2 × 104 × (NB )0.2
Where, R is the cumulative weight of the drop retained
where, where FG is volumetric flow rate of the gas,
in the sieve of diameter d and d′ is characteristic dro-
Nm3/s. The data from He and Standish (Standish
plet size related to 36.8% weight is retained. The expo-
and He 1989; He and Standish 1990) were also used
nent n is a constant and is dependent only on the
in the correlation. Figure 10 shows the effect of top
homogeneity of the drop sizes. One can instead
gas flow rate on the droplet generation rate (g/sec).
define a limiting maximum size for which retained
He and Standish (1990) had further shown that simul-
weight R is 0.1%. Then the distribution can be written
taneous bottom gas injection can increase droplet gen-
as:
eration especially when they were nearly coaxial with
the top jet. These results were obtained from model 1.26
R = 100(0.001)(d/dlimit ) .
studies with subsonic jets of air and at room tempera-
ture. Moreover, there was no slag phase, presence of dlimit needs to be fitted into the actual plant measure-
which can change the rate of generation substantially. ments. Though these approximations and correlations
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 25

are clearly inadequate, most modellers use these for molten oxidising slags have shown that the drop gets
lack of better correlations. buoyed up to the surface as soon as decarburisation
starts, and stays at the surface till CO bubbling sub-
sides. For example, Min and Fruehan (1992) have
Droplets in the emulsion
investigated a single droplet with X-ray fluoroscopy
Kozakovitch (1969) found large amounts of metallic and confirmed an existence of a gas halo around the
droplets in the foamy slag formed during high phos- metal droplet during fast reaction period, the droplet
phorous iron refining. Meyer et al. (1968) made similar remaining at the surface of the slag. They also
observations by collecting samples ejected through the confirmed that the droplet remains largely intact
tap hole in a 230 T BOF converter and analyzing them. though the pictures become hazy when the gas halo
Many investigators have also made similar observation. is present. The pictures show that droplet residence
As mentioned earlier the droplets were in various states time is dependent on bubble formation which keeps
of advanced refining, some of them being almost com- the droplet afloat.
pletely refined, though the bath still had considerable There are two views on how the CO formation keeps
amount of carbon. The fraction of metal in the emul- the drop buoyant. Dogan and coworkers (Dogan et al.
sion were estimated to be large, being almost 25% of 2011c) have formulated a bloated drop theory wherein
the bath weight. This would correspond to a surface CO forms homogeneously inside the drop and this hol-
area of about 40000 m2 if one assumes an average low drop has a low apparent density, due to which it
size of 1 mm for the droplets. Meyer et al. (1964) pro- remains afloat. The other view is that the bubbles
posed that refining in the BOF takes place primarily in form heterogeneously at the drop/slag interface and
the emulsion phase, the bath seeing refining by dilution as long as the bubbles stay attached to the drop they
from droplets falling back (emulsion theory). Emulsion keep it afloat (Sarkar et al. 2015). The visual evidences
in the BOF language refers to a slag-metal-gas system. from X-ray fluoroscopic studies cannot clearly dis-
One can visualise it as a slag-gas foam in which metal tinguish between these two. The fact that there does
droplets are distributed. not seem to be a nucleation barrier during vigorous
Meyer et al. (1968) also reported that many of the deoxidation as evidenced by copious evolution of
droplets displayed high oxygen super-saturation and bubbles suggests interface nucleation. At high carbon
postulated that the finer droplets might have been gen- concentrations when carbon mass transfer within the
erated by homogeneous nucleation of CO and droplet drop is not rate controlling, the highest CO supersa-
bursting. Some droplets showed evidences of being turation should be seen at the drop surface. One there-
attached to gas bubbles and some were even hollow. fore expects the nucleation to take place
There were many experiments with magnetically heterogeneously at the surface, the bubble spending
levitated and freely falling droplets reacting with oxi- some time at the interface before detaching. Since
dising gases (Baker et al. 1964; Robertson and Jenkins there may be several bubbles attached, the drop
1970; Roddis 1973; See and Warner 1973). The results remains buoyed. As carbon falls to low values, nuclea-
were interesting. When the carbon content is high one tion at the interface becomes sporadic, and in periods
would see reactions taking place at the surface, as evi- when there is no bubble attached, oxygen dissolves
denced by CO burning. As the carbon contents come into the metal and diffuses in. Thus the highest super-
down, small droplets are thrown out indicating subsur- saturation region moves inward, first to subsurface and
face nucleation. Further lower in carbon content, the then to deep inside the drop. One may therefore see
droplets sometimes burst, indicating oxygen super-sat- subsurface nucleation initially throwing out small dro-
uration and nucleation deep within the droplet. Super- plets and then deep inside, as seen by El-Kaddah and
saturation to the extent of about 50 atmosphere (for Robertson (1977). These homogeneous nucleation
equilibrium CO) had been reported at the time of dro- events are probably sporadic, with a stochastic nature.
plet bursting (El Kaddah and Robertson 1977, 1978). Simultaneously the apparent density of the drop with
Price (1974) measured the residence time of the dro- no or few bubbles is now high and it falls down into
plets in a BOF vessel by radioactive gold isotope tracer the metal bath. The critical carbon content when the
technique. The maximum residence time of droplets droplet falls down would depend on droplet size, the
which are in advanced state of decarburisation has oxidising potential of the slag (and the rate of mass
been estimated to be about 2 min. Residence time cal- transfer of FeO) and the sporadic nucleation event
culated on the basis of free fall is of the order of a few either at the surface or inside the droplet. Empirical
seconds even while considering the slag to be emul- work to correctly predict the critical carbon content
sified to much greater height. The high residence is lacking. Evidence from levitated drop experiments
time therefore needs an explanation. too point towards these series of events, though the
Several experiments (Min and Fruehan 1992; Mollo- stirring due to the electro-magnetic field makes the
seau and Fruehan 2002; Coley 2013) using X-rays for condition different from that in the BOF especially
visualisation of a single iron-carbon drop reacting in with respect to mass transfer within the drop.
26 G. SNIGDHA ET AL.

Mass transfer in the emulsion consequence on the reaction dynamics, since the
metal droplets are removed from the top layer and
Rates of mass transfer ṅ′′i to droplets in the emulsion is
the refined droplets from the emulsion fall back at
normally given as:
the top. Since much of the heat is also released in the
ṅ′′i = A.km,i .(Cib − Cii ) slag, the slag and the droplets falling back are also hot-
ter. There can also be composition and temperature
where A is the surface area, km,i is the mass transfer stratification due to the scrap at the bottom slowly dis-
coefficient for the species i and Cbi , and Cii are concen- solving into the metal.
tration of the species in the bulk slag and at the inter- High mixing times also correspond to high irrepro-
face respectively. Mass transfer coefficients in the ducibility in mixing times, leading to irreproducible
continuous phase, in dimensionless correlation form, blow behaviour in the absence of bottom blowing.
on droplets where the interface is vigorously stirred For example, a large eddy of metal containing higher
by bubble formation and detachment, by carefully carbon from the bottom being brought to the surface
crafted direct measurements are not available. For of the metal bath can suddenly increase the rate of dec-
solutes in the metal phase, one can make an approxi- arburisation leading to instabilities. Inert gas injection
mation based on Higbie’s surface renewal theory from the bottom of the converters to bring down the
(Dogan et al. 2011b, p 1093): mixing time, therefore, has become standard practice.
 Since the rate of bottom gas injection and the position
2 Di .ud
km,i = . of the porous elements through which the gas is intro-
p rd
duced have a bearing on the reaction dynamics, it is
where Di is the diffusivity of the solute, ud is the vel- necessary to quantify the mixing behaviour for quanti-
ocity of the drop with respect to the fluid outside and tative predictions of composition and temperature
rd is its radius. This is based on the assumption that evolution. A single average t95 value is inadequate for
the liquid inside the droplet recirculates with no signifi- incorporation into a comprehensive model of the
cant viscous resistance. For very small drops, on the BOF since two different mixing curves can give the
other hand, one may make a rigid drop approximation, similar t90 and different t95 values (Figure 11). The
and estimate the mass transfer coefficient to be of the compromise can therefore be a two parameter model,
order of Di/(rd/3). In the slag phase one often takes a based on estimation of two mixing times, say t90 and
value in the range of 10−5–10−4 m/s as an estimate, t95. One can then idealise the metal bath as consisting
for lack of data. In the context of BOF this is clearly of two stirred tank reactors (STR), exchanging metal
inadequate, since the droplet surface is continuously continuously (Krishnakumar et al. 1999). The bottom
disturbed by bubbles. This has two counteracting part will see only scrap melting and the top part will
effects. Part of the drop surface is covered by the gas see all other phenomena explained earlier. The two
bubble and is not available for mass transfer from the parameters of this model, ratio of reactor sizes and
slag to the drop. The drop surface is also vigorously the metal exchange rate, can then be fitted to the mix-
stirred by the formation and detachment of bubbles, ing times of the converter under various conditions of
enhancing mass transfer locally. Several indirect esti- operation.
mations have been made. For example, Gu et al.
(2016) made indirect estimation of mass transfer coeffi-
cient for FeO in slag for phosphorus transfer rate in Lime dissolution
high temperature single drop experiments, and got The BOF needs to maintain a good fluid slag of high
values between 10−5 and 10−4 m/s. Molloseau and basicity (high CaO content) so that the large amount
Fruehan (2002) estimated similar values. Proper exper- of CO generated can be handled, and P can be removed
imental studies, both in cold and hot models, are efficiently. Operators therefore try to obtain a CaO/
necessary to get reasonable correlations in terms of SiO2 ratio in excess of 3.0 in the final slag. Figure 12
dimensionless variables. (a) shows the liquidus contours in a CaO-SiO2-FeO
ternary diagram (Verein Deutscher Eisen.(VDEh).
1995 or Slag Atlas). It is clearly seen that a CaO/SiO2
Mixing within the metal bath
ratio that one can achieve in this system at 1350°C,
Though the slag is very well stirred in the BOF due to i.e. at the beginning of a blow is limited to about 1.6–
the gas jet and a large amount of gas passing through it, 1.7. Marginal improvement may take place with MgO
the metal bath in top blown converters is compara- additions (Figure 12(b)) and some Al2O3 coming
tively poorly. Measurement of mixing time (t95) in from the carry over slag (Verein 1995). In the final
top blown BOF can be as high as 150–180 s, as com- slag too at 1650°C with 25–30% FeO, the maximum
pared to 10–20 s in bottom blown OBM converters. CaO/SiO2 remains less than 3.0. This is also borne
Direct evidence of poor mixing was provided by Chat- out by the slag analyses which often show un-dissolved
terjee et al. (1976) as explained earlier. This has lime. Apart from the issue of solubility of CaO in BOF
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 27

Figure 11. Two schematic mixing curves with same t90 and different t95 values.

Figure 12. Phase Diagram (a) Liquidus isotherms of CaO-SiO2-FeO system (b) Effect of MgO addition on the liquidus (Scheel 1975).
28 G. SNIGDHA ET AL.

slags, the lime particles get passivated in the presence of the oxygen jet and the impact region, and the slag
highly siliceous slags. Since the CaO concentration is region which mostly is in the form of a slag-metal-
the highest at the surface of a dissolving lime particle, gas emulsion.
dicalcium silicate forms here. This compound is not There are three distinct regimes in the blow. The
only highly refractory, it forms an adherent layer initial part is characterised by a bare metal bath covered
retarding further dissolution (Derge 1967). A lime par- with islands of solid lime and some slag carried over
ticle remaining undissolved for long at the high temp- from the previous heat (Figure 13). Jet of oxygen hits
eratures also sinters and becomes less reactive. One the metal bath and does two primary things. First, it
way of breaking the adherent layer is to have high oxidises almost an entire layer of the metal giving
FeO content in the slag. This is the reason for the prac- CO, SiO2, MnO, P2O5 and lots of FeO as discussed ear-
tice of raised lance blowing in the first few minutes of lier. Not all oxygen is consumed in this location, and
the blow, when the FeO is built up to 25–35% or higher. the gas above therefore may contain high ratio of
Though the effect on solubility of CaO is marginal CO2/CO and some oxygen as seen in exhaust gas
(Figure 12), this facilitates breaking of the adherent analysis (Jalkanen and Holappa 2014). The jet also
dicalcium silicate layer permitting further dissolution. throws droplets into the gas phase, which after free
Additives like fluorspar (CaF2) can bring about this flight fall back. Since the gas is oxidising, the droplets
effect much more efficiently, though this is not an get refined during the flight. At the surface of the
acceptable plant practice in recent times for various drops, the order of reactions should be dictated by
reasons. thermodynamics.
For each of the solutes, reaction will involve mass
transfer steps: mass transfer of CO2/O2 in the gas
Process flow and reaction dynamics
phase and of the solutes in the liquid phase. The inter-
With this brief review of the several phenomena of facial chemical reactions are expected to be very fast at
interest in the BOF, we can now take a comprehensive this temperature. The order of the reactions can be
look at the BOF process dynamics. The contents of the obtained by solving the mass transfer equations along
BOF can be divided into several important regions: the with free energy minimisation for the interface reac-
metal bath, which itself can be divided into the bottom tions competing for oxygen. The order is normally Si
and top part between which there is exchange of metal, and Mn followed by C and P. Since the time of flight

Figure 13. Reaction sites in the BOF during the initial period (no slag layer).
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 29

is typically of the order of a second or lower, the dro- decarburisation reaction starts. To maximise the refin-
plets fall back probably completing only part of the sili- ing therefore one should quickly reach a stage where
con reaction. Smaller the drop, further will the refining the decarburisation reaction starts before the drops
proceed due to larger specific surface area. Reaction at fall back. One way to accelerate the reactions is to
the rest of the surface of the metal bath should be small keep the FeO content in the slag high. Another reason
due to smaller surface area compared to that of the dro- why FeO should be increased as early as possible is to
plets. The mass transfer in the gas phase can easily be have a fluid slag by the time decarburisation rate
calculated by Ranz-Marshall type correlations (Bird reaches its highest value, since the large amount of
et al. 1960). At this initial phase of the process droplets gases should quickly escape from the slag. Else, the
are high in solutes, and the gas phase mass transfer is emulsion height will build up uncontrollably leading
expected to be rate controlling. The small drops can to overflow, and slopping.
be considered as rigid and one can assume pure diffu- The FeO content in the slag is a balance between its
sion of solutes inside the droplets. The mass transfer generation at the point of impact and its consumption
coefficient then can be of the order of Di/(rd/3), by the droplets in the emulsion. The FeO generation is
where Di is the diffusivity and rd is the droplet radius. probably weakly dependent of the lance height,
When the drops fall back the condensed phase whereas a high lance leads to less droplet generation
oxide products in the droplets remain at the top of due to lower force with which the jet strikes the
the bath, and on combining with the oxides from metal bath, and vice versa. Hence a raised lance prac-
the impact site and the fluxes added start forming tice, called the soft blow, leads to quick increase in
a molten slag. As mentioned earlier good amount the FeO content in the slag. This facilitates lime dissol-
of FeO is formed at the impact site, and therefore ution and formation of a fluid slag. The initial soft
liquid slag formation should be easy. After some blow, normally 3–4 min, is the normal plant practice.
time, there will be a liquid slag layer covering the At an optimum moment the lance is lowered to
metal bath. Increasingly more and more drops will induce high rates of reactions. Droplet generation
now be thrown to the slag. The drops ejected into rates are high, the bath is already desiliconized hence
the gas phase now have to pass through the slag the droplets undergo vigorous decarburisation till car-
phase before reaching the metal bath. Further refin- bon goes to low values before falling back. During this
ing therefore takes place in the slag. Initially when period of peak decarburisation rates, therefore a large
the slag layer is thin and the drops are high enough part of the metal bath remains in the emulsion as dro-
in silicon and manganese, the drops will fall through plets. These droplets have spent various times in the
before the carbon reaction starts, viz., with no gas emulsion and therefore are in various stages of refine-
evolution, especially for larger drops. Smaller drops ment. The degree of refinement also depends on the
high in carbon may however start to decarburise droplet size. The droplet are therefore characterised
early releasing CO, and slowly emulsifying the slag. by two variables; the time it was formed (and therefore
This early phase is characterised by a low flame at its age) and the size of the droplet. The droplet starts to
the mouth, since CO formation is comparatively fall back when a characteristic carbon content is
low. Once the silicon in the metal bath falls to reached, which depends on its size, the slag FeO and
some extent, the desiliconization progresses consider- temperature. During this last phase of the droplets,
ably before the droplet has fallen down. Carbon reac- the oxygen potential at the interface is also high and
tion starts and the droplet now stays buoyed in the therefore phosphorus is also removed if other con-
emulsion till its carbon content reaches the critical ditions are favourable. Falling drops results in apparent
carbon content as explained earlier. In the slag refinement of the top of the metal bath, which on mix-
phase, the rate is expected to be controlled by slag ing lead to refinement of the rest of the bath. Since the
phase mass transfer of FeO, as long as carbon in time for refinement of a drop may be of the order of ½–
the drop remains high enough. Once a critical carbon 2 min, one sees drop in C, Mn and P in the bulk metal
is reached in the droplet, bubbling slows down and sample even if Si in the sample is still significant
then ceases, and the drop falls down. The critical car- (Figure 2).
bon is largely determined by the FeO content in the The overall rate of reaction, to some extent, is self-
bath. Quickly the emulsion builds up and we enter correcting. If the number of droplets in the emulsion
the second phase of reactions in the emulsion. The come down decreasing the rate, the level of the metal
flame at the mouth becomes large. The lance tip bath increases leading to lower effective lance distance,
gets dipped into the emulsion. which in turn causes droplet generation to increase
In the second phase almost all of the droplets are (Meyer et al. 1964). This is one of the reasons for the
ejected into the emulsion, and the gas phase reactions near constant decarburisation rate during middle part
become unimportant. One should note that the resi- of the blow. One should however note that the C con-
dence times of the quiescent droplets in the slag are tent of droplets entering into the emulsion keeps falling
only of the order of a few seconds unless down, and therefore their residence time. One needs to
30 G. SNIGDHA ET AL.

correspondingly increase the droplet generation rate by equation, one instead writes an equation for a phos-
progressively lowering the lance. Figure 14 indicates phate capacity CPO3−
4
of slag as
the lance height variation during a typical blow (Cicutti
⎛ ⎞
et al. 2000). 1 5
As the bath itself becomes low in carbon in the final ⎜ ⎟
CPO3−
4
= (%P)slag /⎝ p2p2 . p4O2 ⎠,
phase of the blow, the rate now gets limited by carbon
diffusion within the droplet even as it enters the emul-
sion. The CO generation is low and is not able to keep
and correlates the phosphate capacity to the slag com-
the droplets floated. The residence time drops down to
position empirically. Both these approaches are con-
a few seconds, and therefore number of droplets in the
ceptually similar. The partial pressures of P2 and O2
emulsion comes down even though the lance height
can easily be converted to wt% dissolved in metal or
has been brought down to the lowest level permissible
activity of FeO in slag with known thermodynamic
for lance health. The emulsion dies down. One sees at
data. The slag data as a function of composition either
this time falling rates of decarburisation and fast
as γP2O5 or as CPO3− have been empirically determined
buildup of FeO in the slag. Since the rates are low, 4
by several researchers. The progress of partition coeffi-
the FeO content and the oxygen dissolved in steel
cient LP for P between bulk slag and bulk metal can be
increase much beyond what is dictated by C-O equili-
calculated when the slag analysis during the blow is
brium. At this period one therefore raises the argon
known as has been done for example by Basu et al.
stirring rate, increasing thereby the droplet generation
(2007) (Figure 15). Similar calculations have been
rate without adding extra oxygen. This helps to some
done, for example, by Diao et al. (2015) and others.
extent.
In the initial times of the blow, the bath carbon is
quite high and contains silicon too. Therefore at the
Phosphorus reaction slag/metal bath interface, the oxygen potential remains
low. One therefore does not expect very high rates of
Phosphorus removal is sometimes a problem in the
dephosphorization at the bulk metal/slag interface.
BOF operation and may result in re-blows, especially
The metal droplets on the other hand get highly
when the input phosphorus in the hot metal is high
refined in a matter of 1–2 min, and before returning
(∼0.2%). Though the conditions are generally favour-
to the metal bath have high oxygen potential at its
able in the final slag with high FeO and high basicity,
interface. Further the partition coefficient LP at this
one can land in adverse situation if the slag regime is
time is high since FeO content is high due to soft
not carefully managed throughout the blow.
blow, temperature is low, albeit with some CaO yet
The thermodynamics of phosphorus is well known.
to dissolve. The number of droplets in the emulsion
One can write the reaction either in terms of molecular
is also very large. Therefore the dephosphorization
species or in the ionic form:
rate is very high as can be seen in Figure 2. Towards
5 the end of the blow again, the conditions in the slag
P2 + O2 (g) = P2 O5 (l)
2 are favourable with high FeO and high basicity, though
1 5 3 now the temperature would have risen substantially.
P2 + O2 (g) + O2− = (PO3−
4 ) The rate of phosphorus removal however is not very
2 4 2
high in this period, since the number of droplets is
In the former case one writes an equilibrium con-
stant as K = γP2O5.XP2O5/(p2P2.p2.5
O2), and expresses the
Raoultian activity coefficient, γP2O5, as a function of
slag composition. If one adopts an ionic form of the

Figure 15. The progress of partition coefficient P between bulk


Figure 14. Lance height variation (Cicutti et al. 2000). slag and bulk metal during the blow (Basu et al. 2007).
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 31

not very high, surface area is quite small and therefore dynamics. The model by Pahlevani et al. (2010) has
all reactions are slow. Vigorous argon stirring is helpful kinetic approach to the reaction; however, the detailed
at this time of the blow, and for some time after the description of the reaction sites and their contribution
oxygen flow is stopped, though to a limited extent. is lacking. The oxygen has been partitioned in the ratio
It is in the middle part of the blow one has the great- of 20:70: 10 between carbon, iron and post combustion
est opportunity for efficient overall dephosphorization. of CO to CO2 at the impact site; the FeO then reacts
After the soft blow when the lance is lowered progress- with solutes in the metal. The model by Kruskopf
ively for effecting high rates of decarburisation, FeO and Visuri (2017) calculates the reaction progress by
content in the slag significantly drops and remains free energy minimisation to partition the oxygen
low till the emulsion starts collapsing. The slag between various elements; it however considers a single
becomes comparatively ‘dry.’ The LP becomes adverse, reaction zone in a lumped fashion.
and one can easily get phosphorus reversal to the metal Models by Dogan et al. (2011a) and Sarkar et al.
as can be seen in Figure 3. Lower the FeO level greater (2015) incorporate most of the details of the dynamics
the reversal. This reversal increases the load on the last discussed earlier. Dogan et al. (2011b) divided the pro-
part of the blow where the rates of reactions are anyway cess into two conceptual reactors: the metal bath and
low as explained earlier. the slag-metal emulsion. The reactions are primarily
Close control of the FeO content during the middle at the impact zone (Dogan et al. 2011a, p. 1086) in
part of the blow is therefore necessary if one desires to the metal bath and at droplets distributed in the emul-
make low phosphorus steel. Premature lowering of the sion (Dogan et al. 2011b, p. 1093). Aspects like droplet
lance in each stage can lead to very low FeO content (< generation rates, droplet residence times were included
12–15%). FeO content is determined by the balance in the model description; reactions of other solutes like
between droplet generation rate (consumption rate) Si and Mn are however not included. Overall the model
and the FeO generation rate as explained earlier. One is quite comprehensive. Sarkar et al. (2015) divided the
should also note that very low FeO in the slag lowers furnace into three well-stirred reactors, the metal bath
decarburisation rate too. Very high FeO on the other being divided into two to account for mixing times.
hand leads to sloppy conditions. Solutes like C, Si and Mn are taken into account, the
One can achieve higher levels of FeO content by oxygen partition being calculated by mass transfer
modifying the lance practice. The lance height for the steps as well as free energy minimisation at the interface
intermediate levels can be kept slightly higher than of droplets. Both these models use a single average dro-
the normal. One can also slightly delay lowering of plet size and assume a temperature evolution pattern.
the lance, taking care to see that it does not lead to Further developments in the model of Sarkar et al.
uncontrolled emulsion build up. One can achieve this (2015) incorporating droplet size distribution using
by distributed ore addition during this period too RRS distribution and tracking the age of droplets of
(Ogasawara et al. 2013). each size have been made (Ghosh 2014; Biswas 2015).
Improvement in quantitative understanding of
many of the phenomena like those at the impact site
Dynamic modelling
and at bubble stirred droplets is expected to lead to
Since an extensive review of dynamic models is beyond models that can be used for control and optimisation
the scope of this paper, a brief review of the more with greater confidence.
recent attempts is presented here.
Though the dynamic modelling efforts had started
fairly early in the history of the BOF (Asai and
Summary
Muchi 1970), enhanced emphasis on model based con- The BOF process is a process of very high kinetics, the
trol has led to renewed interest in the subject. Due to reactions taking place at multiple locations. The jet-
lack of good data for quantitative description and liquid interaction and the carbon oxygen reaction gen-
some lack of clarity in the understanding of the reac- erating a gaseous product have enormous effects on the
tion dynamics itself, most of these models have adjus- overall process dynamics. Several sub-processes need
table parameters which can make the predictions to be well understood, quantitatively, for a full descrip-
match with plant data fairly well. Many models have tion of the process. Some of these processes still need
concentrated on one or the other aspects of the process further work to put models for the process on a firm
dynamics. Shukla et al. (2010), for example, have footing. This paper has attempted to present a compre-
adopted a thermodynamic approach, distributing the hensive view of the process dynamics with the present
oxygen among solutes based on the Gibbs energy of understanding of the sub-processes.
oxidation without consideration of the kinetic aspects.
Model by Kattenbelt and Roffel (2008) is based on their
experiments on step response to process variables, and Disclosure statement
uses simple empirical kinetic equations to model the No potential conflict of interest was reported by the authors.
32 G. SNIGDHA ET AL.

ORCID Co-Co2 gas mixtures at high pressures. Metall Trans B.


9(2):101–199.
Ballal N. Bharath http://orcid.org/0000-0001-6966-0021 Ghosh S. 2014. Dynamic modeling of LD converter steel-
Nurni N. Viswanathan http://orcid.org/0000-0003-0278- making [M Tech thesis]. IIT Bombay.
1333 Gu K, Dogan N, Coley KS. 2016. Proceedings of the 10th
International Conference on Molten Slags, Fluxes and
References Salts, R. G. Reddy et al, Ed..
He QL, Standish N. 1990. A model study of residence time of
Abramovich GN, Girshovich TA, Krasheninnikov SI, metal droplets in the slag in BOF steelmaking. ISIJ Int. 30
Sekundov AN, Smirnova IP. 1984. The theory of turbulent (5):356–361.
jets. Moscow Izd: Nauk. Jalkanen H, Holappa L. 2014. Converter steelmaking.
Alam M, Naser J, Brooks G. 2010. Computational fluid Treatise on process metallurgy. Seetharamn et al ed.
dynamics simulation of supersonic oxygen jet behavior Elsevier. 3A:223–270.
at steelmaking temperature. Metall Trans B. 41:636–645. Kattenbelt C, Roffel C. 2008. Dynamic modeling of the main
Anderson AR, Johns FR. 1955. Characteristics of free super- blow in basic oxygen steelmaking using measured step
sonic jets exhausting into quiescent air. J Jet Propuls. responses. Metall Mater Trans B. 39:764–769.
25:13–15. Koria SC, Lange KW. 1984. A new approach to investigate
Asai S, Muchi I. 1970. Theoretical analysis by the use of the drop size distribution in basic oxygen steelmaking.
mathematical model in LD converter operation. Trans Met Trans B. 15(March):109–116.
ISIJ. 33:25–63. Kozakovitch P. 1969. Foams and emulsions in steelmaking. J
Baker LA, Warner NA, Jenkins AE. 1964. Kinetics of decar- Metals. 22(7):57–68.
burization of liquid iron in oxidizing atmosphere using Krishnakumar K, Ballal NB, Sardar MK, Sinha PK, Jha KN.
levitation technique. Trans TMS-AIME. 230:1228–1235. 1999. Water model experiments on mixing phenomena in
Basu S, Lahiri K, Seetharaman S, Halder J. 2007. Change in a VOD ladle. ISIJ Int. 39(1):419–425.
phosphorus partition during blowing in a commercial Kruskopf A, Visuri V-V. 2017. A gibbs energy minimiz-
BOF. ISIJ Int. 47(5):766–768. ation approach for modeling of chemical reactions in a
Bird RB, Stewart WE, Lightfoot EN. 1960. Transport basic oxygen furnace. Metall Mater Trans B. 48:3281–
phenomena. NY: John Wiley. p.409, 1960. 3300.
Biswas J. 2015. Dynamic modeling of LD converter steel- Lee YE, Kolbeinsen L. 2007. An analysis of hot spot phenom-
making [M Tech thesis]. IIT Bombay. enon in BOF process. ISIJ Int. 47(5):764–765.
Bradshaw AV, Richardson FD. 1970. Chemical engineering Li K, Dukelow DA, Smith GC. 1964. Decarborization of iron
in the iron and steel Industry. Symposium, London, carbon system in oxygen steelmaking. Trans AIME. 230
Institution of Chemical Engineering. (July):71.
Chatterjee A, Lindfors MA, Wester JA. 1976. Process metal- Meyer HW, Dukelow DA, Fischer MM. 1964. Static and
lurgy of LD steelmaking. Ironmaking Steelmaking. 3 dynamic control of the basic oxygen process. J Met. 16
(1):21–32. (6):501–507.
Cicutti C, Valdez M, Pérez T, Donayo R, Petroni J. 2002. Meyer HW, Porter WF, Smith GC, Szekely J. 1968. Slag-
Analysis of slag foaming during the operation of an indus- metal emulsions and their importance in BOF steelmak-
trial converter. Lat Am Appl Res. 32(3):237–240. ing. J Metals. 25:35–42.
Cicutti C, Valdez M, Pérez T, Petroni J, Gomez A, Donayo R, Miller TW, Jimenez J, Sharan A, Goldstein DA. 1998.
Ferro L. 2000. Study of slag-metal reactions in an LD-LBE Oxygen steel making process. The making, shaping and
converter. 6th int. conf. molten slags; Fluxes Salts, 367. treating of steel, 11th Edition, Vol.2, Chap.9, 475–524.
Coley KS. 2013. Progress in the kinetics of slag-metal-gas Min DJ, Fruehan RJ. 1992. Rate of reduction of feO in slag by
reactions, past, present and future. J Min Metall Sect B Fe-C drops. Metall Trans B. 23(Feb.):29–37.
Metall. 49(2):191–199. Molloseau CL, Fruehan RJ. 2002. The reaction behavior of
Derge G. 1967. Transactions TMS-AIME. 239:1483–1487. Fe-C-S droplets in CaO-SiO2-MgO-FeO slags. Metall
Diao J, Qiao Y, Liu X, Zhang X, Qiu X, Xie B. 2015. Slag for- Mater Trans B. 33(3):335–344.
mation path during dephosphorization process in a con- Molloy NA. 1970. Impinging jet flow in a 2-phase system-
verter. Int J Miner Metall Mater. 22:1260–1265. basic flow pattern. J Iron Steel Inst. 208:943.
Dogan N, Brooks GA, Rhamdhani MA. 2011a. Odenthal H, Falkenreck U, Schlüter J. 2006. CFD simulation
Comprehensive model of oxygen steelmaking part 1: of multiphase melt flows in steelmaking converters. Eur
model development and validation. ISIJ Int. 51(7):1086– Conf Comput Fluid Dyn. 1–21.
1092. Ogasawara Y, Miki Y, Uchida Y, Kikuchi N. 2013.
Dogan N, Brooks GA, Rhamdhani MA. 2011b. Development of high efficiency dephosphorization system
Comprehensive model of oxygen steelmaking part 2: in decarburization converter utilizing FetO dynamic con-
application of bloated droplet theory for decarburization trol. ISIJ Int. 53(10):1786–1793.
in emulsion zone. ISIJ Int. 51(7):1093–1101. Orban AR, Hummell JD, Cocks GG. 1961. Research on con-
Dogan N, Brooks GA, Rhamdhani MA. 2011c. trol of emissions from bessemer converters. J Air Pollut
Comprehensive model of oxygen steelmaking part 3: dec- Control Assoc. 11(3):103–113.
arburization in impact zone. ISIJ Int. 51(7):1102–1109. Pahlevani F, Kitamura S, Shibata H, Maruoka N. 2010.
El-Kaddah NH, Robertson DGC. 1977. Homogeneous Simulation of steel refining process in converter. Steel
nucleation of carbon monoxide bubbles in iron drops. J Res Int. 81(8):617–622.
Colloid Interface Sci. 60(2):349–360. Price DJ. 1974. L.D. steelmaking: significance of the emul-
El-Kaddah NH, Robertson DGC. 1978. The kinetics of gas- sion in carbon removal. In: Jones MJ, editor. Process
liquid metal reactions involving levitated drops. engineering pvrometallurgy. London: The Inst. Min.
Carburization and decarburization of molten iron in Metall.; p. 8–15.
MINERAL PROCESSING AND EXTRACTIVE METALLURGY 33

Robertson DGC, Jenkins AE. 1970. Hetrogeneous kinetics Standish N, He QL. 1989. Drop generation due to an imping-
at elevated temperatres. New York: Plenum Press; ing jet and the effect of bottom blowing in the steelmaking
p. 393–408. vessel. ISIJ Int. 29(6):455–461.
Roddis PG. 1973. Mechanism of decarburization of Fe-C Subagyo GA, Brooks KS, Coley KS, Irons GA. 2003.
alloy drops falling through an oxidizing gas. J Iron Steel Generation of droplets in slag-metal emulsions through
Inst. 211:53–58. top gas blowing. ISIJ Int. 43(7):983–989.
Sarkar R, Gupta P, Basu S, Ballal NB. 2015. Dynamic model- Szekely J, Themelis NJ. 1971. Rate phenomena in process
ing of LD converter steelmaking: reaction modeling using metallurgy. New York: John Wiley Sons Inc. 784.
Gibbs’ free energy minimization. Metall Mater Trans B. 46 Turkdogan ET. 1996. Fundamentals of steelmaking. 1st ed.
(2):961–976. London: The Institute of Materials.
Scheel R. 1975. Sprechsaal Keram., Glas. Baeustoffe. 108(23- Turner G, Jahanshahi S. 1987. A model investigation on emul-
24):685–686. sification of metal droplets in the basic oxygen steelmaking
See JB, Warner NA. 1973. Reactions of Fe alloy drops in processes. Trans Iron Steel Inst Japan. 27(9):734–739.
free fall through oxidizing gases. J Iron Steel Inst. Verein Deutscher Eisen.(VDEh). 1995. Slag atlas. 2nd Ed.
211:44–52. Verlag StahlEisen; p 223.
Shukla AK, Deo B, Millman S, Snoeijer B, Overbosch A, World Steel Association. 2017. Steel statistical yearbook.
Kapilashrami A. 2010. Steel Res. Int. 81:941–948. 1–128.

Das könnte Ihnen auch gefallen