Oden Thal 2010

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Simulation of Fluid Flow and Oscillation of the Argon Oxygen

Decarburization (AOD) Process


HANS-JUERGEN ODENTHAL, UWE THIEDEMANN, UDO FALKENRECK,
and JOCHEN SCHLUETER
The oscillation of argon oxygen decarburization (AOD) converters is ow related and depends
on the process parameters (e.g., vessel geometry, melt ll height, process gas type and blowing
rate, vessel tilting angle, as well as geometry, number, and arrangement of the side-wall nozzles).
For a 120-ton AOD converter with seven submerged side-wall nozzles, plant tests, physical
simulations on a 1:4 scale water model, and computational uid dynamics simulations have
been done. The investigations show that the penetration depth of an inert gas jet into the melt
does not exceed approximately 0.4 m. The plumes are located close to the nozzle-side converter
wall and induce a large-scale primary vortex as well as intensive surface movements; both are
responsible for the oscillation. Several process mechanisms were investigated. The oscillation is
highest in the last stage of the dynamic blow and is still high during the reduction stage. As the
amount of inert gas increases, the vibration level also increases. Inert gas has a greater inuence
on the oscillation than oxygen. Tilting the converter around 8 deg clearly leads to more
intensive oscillations. Increasing the blowing rate increases the forces and torques acting on the
vessel, whereas the oscillation frequency remains nearly constant. A varying ll level does not
inuence the vibration level the same way as the blowing rate. The operational test shows, for
example, that the maximum torque does not depend on the heat size when the latter varies
between 8 pct and +21 pct of the nominal heat size. The water model test shows decreasing
forces and torques with a rising ll level.
DOI: 10.1007/s11663-009-9335-y
 The Minerals, Metals & Materials Society and ASM International 2010

I.

INTRODUCTION

IN 2008, around 26 million tons of stainless steel


were produced around the world. Stainless steel contains
a minimum of 10 pct chromium and various amounts of
Ni, Mo, Mn, and other elements to arrive at the desired
material properties. The corrosion resistance is caused
by a thin layer of passive, stable chromic oxide, which
reacts slowly. The layer adheres extremely well and,
when combined with oxygen from the air, is self-healing.
To produce stainless steel, a two-stage process is
employed. Scrap and alloying elements, sometimes
together with hot metal, are molten down in the electric
arc furnace. Then the molten metal is rened in the
argon oxygen decarburization (AOD) converter. In the
AOD process, a highly chromium-alloyed melt is
decarburized by injecting and blowing oxygen and inert
gases (N2, Ar) through submerged side-wall nozzles and
a top lance. Depending on the mix of charge materials,
HANS-JUERGEN ODENTHAL, Deputy General Manager,
R&D DivisionFundamentals and Models Melting/SAF, UWE
THIEDEMANN, Deputy General Manager, Steelmaking/Continuous
Casting Technology DivisionProduct Development Steelmaking,
UDO FALKENRECK, General Manager, R&D Division, and
JOCHEN SCHLUETER, Vice President Special Technologies, are
with the SMS Siemag AG, Eduard-Schloemann-Strae 4, 40237
Duesseldorf, Germany. Contact e-mail: hans-juergen.odenthal@
sms-siemag.com
Manuscript submitted May 5, 2009.
Article published online January 26, 2010.
396VOLUME 41B, APRIL 2010

the beginning carbon contents are between 1.5 pct and


4.5 pct, and the chromium contents are between 10 pct
and 25 pct, depending on the steel grade to be produced.
A specic feature in the decarburization of highchromium melts is that chromium itself has a high
anity to oxygen, which requires special process-related
measures to limit its conversion into the slag. The
competitive situation of carbon and chromium in
contact with oxygen is demonstrated by the following
reaction equations:
2C fO2 g ! 2fCOg

1

4Cr 3fO2 g ! 2Cr2 O3

2

where [ ] means it is dissolved in the melt, ( ) means it


is included in the slag, and { } means it is gaseous. At
ambient pressure, for every temperature, an equilibrium exists between the carbon and chromium content
of the melt and the oxygen supplied (Richardson
Ellingham diagram). As the temperature increases, the
free reaction enthalpy DG0 for Eq. [1] decreases,
although it increases for the slagging reaction in
Eq. [2]. Equations [1] and [2] can be combined as follows:
2Cr 3fCOg $ 3C Cr2 O3

3

Equation [3] makes it clear that the equilibrium


between carbon and chromium in the melt depends on
the CO partial pressure in the gas bubbles. To supply
as much oxygen as possible to the carbon while at the
METALLURGICAL AND MATERIALS TRANSACTIONS B

same time minimizing the oxidation of the chromium,


the reaction must proceed toward the left. This
sequence takes place preferably at high-carbon and
low-chromium contents, at high temperatures, and
above all, at a low-CO partial pressure pCO. Increasing
the process temperature is only reasonable up to
1700 C because of excessive refractory wear at higher
temperatures. The critical carbon content that marks
the lower limit where chromium starts to scorify as a
function of pCO is calculated as follows:
v
u
2
2
3

 u
3 K fp
CO g fCr  cpctCr
t
cpctC
4
aCr2 O3 fc 3
where c is the pct per weights of carbon, K is the equilibrium constant, a is the activity, and f is the coecient of activity. The CO partial pressure in the gas
bubble is expressed as follows:


nCO
5
pz
pCO
nCO nAr
where n is the molar mass and pz is the static pressure in
the melt at depth z.
The AOD process uses the dependence of the CrC
equilibrium on the partial pressure by breaking the
process down into various phases and progressively
adding more inert gas to the process gas as the carbon
content decreases. This process reduces the CO partial
pressure so that the decarburization toward low carbon
contents is promoted, while at the same time the
oxidation of chromium is limited. However, for kinetic
reasons, the scorication of chromium cannot be fully
avoided. To reduce the loss of chromium, the input of
oxygen at low-carbon contents of less than, for example,
0.4 pct is, depending on the chromium content of the
melt, increasingly reduced because the achievable decarburization rate here is a function of the melt bath
carbon-content proper and, moreover, chromium scorication increases because of the thermodynamic equilibrium situation.
Figure 1 shows the typical blowing stages of a 120-ton
AOD converter with seven side-wall nozzles. Down to a
carbon content of approximately 0.4 pct, a total of
V_ O2 = 240 m3/minstp is blown for rening, with the
larger proportion of oxygen blown through the singlehole top lance. The decarburization rate is proportional
to the oxygen supplied. The maximum amount of
oxygen is limited by the steel grade, the starting
conditions, the oxygen supply, the waste gas exhausted,
and the slag slopping. At this stage, the low amount of
inert gas through the side-wall nozzles serves for
cooling. In conformity with the laws of kinetics and
thermodynamics at decreasing carbon contents, the
amount of oxygen is reduced in stages, and at the same
time, the process gas is enriched with inert gas (dynamic
blow). The aspect ratio of oxygen and inert gas drops
from the original 8:1 via 1:1 and 1:2 down to 1:3.3, with
the total amount of process gas remaining almost
constant during the dynamic stages. At the end of
decarburization period, the carbon content is roughly
around 0.02 pct for a lot of stainless steel grades.
METALLURGICAL AND MATERIALS TRANSACTIONS B

Decarburization is followed by a pure inert gas


stirring for reduction. During the reduction stage, the
oxidized metalschromium, above allare reduced
mostly by the addition of silicon carriers. In this
process, the initial Cr2O3-containing rening slag
changes from a solid to a liquid state.
Reduction and subsequent deslagging are followed by
a separate desulphurization using a two-slag method.
The basicity of the slag and, thus, the sulphur capacity is
increased by adding lime and uxing elements, with
which simultaneous inert gas stirring causes an almost
complete transition of sulphur from the melt into the
slag. The last inert gas stirring stage (alloying) provides
an opportunity for analysis trimming of the melt before
tapping.
The AOD converter is a metallurgical reactor that
oers excellent mixing conditions because of the high
melt turbulence caused by the injection of large amounts
of process gas through the side-wall nozzles. However,
the AOD process is accompanied by intense converter
vibrations, which from the outside can be recognized by
more or less staggering movements of the vessel around
the axis of rotation. Vessel vibrations act as torques on
the structural components and have to be taken into
account during the design phase, especially with regard
to the dimensioning of the torque support arm.
Figure 1 also shows the vibration intensity throughout the process schematically as a red curve. The carbon
content of the starting melt was around 3 pct. Although
the highest amount of process gas is injected during the
main blow, the vibration intensity during this stage is
relatively low. It is typically already during the main
blowing period that a clear reduction of the vibration
amplitude is to be noticed, which can be attributed to
damping eects caused by the increasing weight of the
slag with high-slag viscosity. Another reduction of
the amplitudes usually is observed toward the end of
the rst blowing stage, which can be explained by a
reduced decarburization rate and the decreasing formation of CO bubbles. Similar phenomena were observed
in vibration measurements on basic oxygen furnace (BOF)

Fig. 1Blowing pattern of the 120-ton AOD converter with characteristic vibration intensity levels.
VOLUME 41B, APRIL 2010397

converters as well.[13] The authors explained this


behavior by reducing the decarburization rate in favor
of an increased iron scorication while approaching the
critical point of decarburization. The general idea was
that strong oscillations were accompanied by an intensive mixing of the melt.
The part the top lance that contributes to the
excitation of vibrations is negligible when compared
with the part supplied by the side-wall nozzles. The
decisive factor in the excitation of vibrations is the
volumetric ow rate of the process gases. However, inert
gas has a much stronger inuence on the vibration than
oxygen. The higher the amount of inert gas, the stronger
the vessel vibrations become. The dierent eect caused
by oxygen and inert gas is explained by the dierent
behavior of these gases immediately after they have
penetrated the melt domain.
When injected into the melt, the oxygen jet immediately collapses, with most of the oxygen rst oxidizing
chromium and then, because of the reduction of Cr2O3
in the presence of overcritical carbon contents, reacting
to form CO bubbles (see Eq. [3]). In contrast, inert gas
bubbles have more time to grow after penetrating the
melt. Consequently, a mechanism must be assumed to
exist in the formation of inert gas bubbles, with
corresponding eects on the bubble size and their
spatial distribution, which is much dierent from the
mechanism existing when injecting pure oxygen. During
the dynamic decarburization stage, the conversion of
oxygen to CO additionally increases the vibration level
induced by inert gas, especially when the decarburization rate is high, and a large amount of oxygen is
converted to CO.
In addition to the amounts of process gas, the
refractory lining degree of wear inuences the vibration.
As the degree of wear increases, the bath geometry and
the center of gravity of the converter vessel change.
Thus, the detected torque shows a maximum at the start
of a converter campaign, which decreases typically by
around 25 pct toward the end of the campaign.
Vibrations are caused by geometry conditions, process gases, rising gas bubbles (N2, Ar, CO), and the
resulting movements of the free surface (i.e., vibrations
are closely linked with uid ow and heat phenomena).
Process variables that inuence the ow conditions are
the converter geometrymainly characterized by the
ratio of the bath diameter and ll levelthe type and
ow rate of process gases, the geometry and alignment
of the side-wall nozzles, and the inclination of the
converter vessel. The interaction of all variables has, to
this date, not been understood in detail. This is because
of both the multitude of variables and the varying
boundary conditions prevailing in a melt shop, which
frequently make it dicult to carry out a systematic
investigation and to achieve reproducible results.
For reliable operation, a long service life of the plant
and its components as well as optimal process conditions and knowledge about the factors inuencing the
vibration behavior are indispensable. Against this background, SMS Siemag, in addition to worldwide vibration analyses on existing AOD converters, also performs
physical simulations on water models and numerical
398VOLUME 41B, APRIL 2010

simulations on computational uid dynamics (CFD)


and the nite element method (FEM). A variety of these
investigations is presented in the following sections.

II.

LITERATURE

Most of the authors carry out investigations on AOD


water model converters,[410] using the modied Froude
number for similarity criterion. The model scales vary
between 1:10[9] and 1:4.[10] Investigations on the penetration length of gas jets into liquids were made by
Bjurstrom et al.,[2] Fabritius et al.,[6] and Tilliander
et al.[9]
Bjurstrom et al.[4] performed laser Doppler anemometry measurements on a water model converter. However, the gas bubbles considerably complicated the
measurements. When the gas ow rates were high, only
a few velocity vectors characterized the ow eld.
According to the authors, the penetration length of the
gas jet depends much more on the blowing rate than on
the ll level. With a low blowing rate, a circulating ow
results in the symmetry plane. As the blowing rate
increases, the area of recirculation is pushed away in the
direction of the wall located opposite the nozzle.
Fabritius et al.[6] investigated the oscillation of the
free-water surface. Regular oscillations occurred when
the gas jet reached up to the center of the bath, and the
heterogeneous buoyant plume uctuated uniformly
back and forth (type A oscillation). In case of a high
blowing rate, the gas jet in the water model reached
beyond the center of the bath, and the main ow was
reversedalthough this result is not likely to occur in
reality. With the same blowing rate, the penetration
length with a small nozzle diameter is higher than with
a large diameter. The dominating frequency during
the decarburization stage of the original converter
(O2/N2 = 1:3) was 1.45 Hz. The frequency was independent of the blowing rate, nozzle diameter, and the
angle between the nozzles.
Fabritius et al.[11] conducted vibration analyses on a
real AOD converter. An acceleration transducer was
installed on one of the side-wall nozzles. The frequency
detected depended only on the height-to-diameter (H/D)
ratio of the melt bath but not on the blowing rate or gas
type. With an increasing H/D ratio, the frequency rose
slightly and was constant for H/D > 0.45. The worn
converter (shallow bath and/or small H/D) induced at
surface waves with a low frequency and long periods,
whereas the freshly relined converter (high bath and/or
high H/D) caused higher frequencies. For a low ll level,
the converter bottom dampened oscillations at the phase
boundary. Vibrations were determined mainly by the
way in which the converter was operated and, to a lesser
extent, by the decarburization rate. The higher the
blowing rate and the proportion of inert gas, the higher
the vibration amplitude (with unchanged frequency).
Likewise, the amplitude increased with a rising ll level.
With a constant ow rate of the process gas, the
injection of inert gas produced more vibrations than the
oxygen injection (i.e., a correlation was observed
between the length of penetration and the vibrations).
METALLURGICAL AND MATERIALS TRANSACTIONS B

The argon jet (nAr = 39 kg/kmol) was longer than that


of oxygen (nO2 = 32 kg/kmol), and consequently, the
amplitudes were higher. According to Fabritius
et al.,[6,11] the vibrations measured for nitrogen (nN2 =
28 kg/kmol) were higher than for argon, which may be
because the maximum amplitude without a root-meansquare (rms) value was considered. Thus, no statistical
evaluation of the measurement was provided. The
authors found the maximum amplitude in the process
stage at V_ O2 N2 146 m3 = minstp and O2/N2 = 1:3, which
coincides with our own operational measurements.
Tilliander et al.[8,9] performed numerical and physical
simulations on the converter and on the nozzle. The
penetration length of the gas jet did not extend to
the center of the bath, and gas bubbles rose close to the
nozzle wall. The mean penetration length of the gas jet
as quoted by Fabritius et al.[6] and Tilliander et al.[9]
was higher than the one indicated by Hoefele and
Brimacombe.[12]
Dierent ow rates, nozzle diameters, angles between
the nozzles, and their inuence on the mixing behavior
were investigated by Fabritius et al.[6] and by Wei
et al.[10] The mixing time was determined by a conductivity measurement after adding saltthis method was
inaccurate because of the inuence of a variable density.
The mixing time decreased with rising blowing rate and
larger angles between the nozzles.
Many vibration analyses also have been made on
BOF converters. Grenfell et al.[1] evaluated the signals
from load cells installed in the bearing pedestals of
130-ton BOF converters. The authors made a correlation between the decarburization rate (i.e., the CO gas
bubble formation rate in the bath and the vibration
amplitudes of the measured signals). This correlation
was used to determine the end-of-blow point based on
the vessel vibration and the carbon content.
Onishi et al.[3] investigated the vibration of a combined blowing 250-ton Kawasaki-basic oxygen process
(K-BOP) converter. They found that a clear reduction of
the vibration amplitudes took place toward the end of
the blowing process because the vibration amplitude
depended on the gas-bubble formation rate in the melt
bath. Because the decarburization rate decreased at the
end of the blowing stage in favor of progressing metal
scorication, fewer gas bubbles were generated, which
led to a reduced melt movement and a lower vibration
level.
Mucciardi et al.[2] seized on the results of Grenfell
et al.[1] and Onishi et al.[3] An oxygen blowing test
performed on a hot-metal charge, and for which an
acceleration transducer was used, conrmed the ndings
of both teams of authors.

III.

AOD CONVERTER

AOD converters today are designed mostly as changevessel units. This design provides for the suspensions
between the vessel and the closed trunnion ring without
water cooling to be released, the process and inert gas
lines to be disconnected, the gas stack and lance system
to be taken to the change position, and the converter
METALLURGICAL AND MATERIALS TRANSACTIONS B

vessel to be lifted out of the trunnion ring by the shop


crane. In contrast, U-shaped, water-cooled trunnion
rings combined with a converter change car are ideal for
converters installed between columns. The trunnion ring
is equipped with a tilt-drive system and hydraulic
equipment.
A. Geometry and Process Data
Table I and Figure 2 show the nomenclature and
simplied rotational symmetric geometry of the 120-ton
AOD converter on which operational tests were performed in the melt shop and for which both physical
simulations on a water model and CFD simulations
were carried out. The diameter-to-height ratio of the
 D1 D2 . The process
 st 1:58 at D
melt bath was D=H
2
gases were injected through seven side-wall nozzles
arranged at the level Hn. Each nozzle consisted of a
coaxial inner tube (Dn,in) for the process gas (O2, N2, Ar)
and a shroud tube (Dn,out) for the cooling gas (N2, Ar).
The nozzles were arranged at an angle of an = 18 deg,
with the horizontal nozzle axes aligned relative to the
vertical centerline of the AOD converter. The distance
between the refractory bottom and the trunnion axis
was Ht. The parameters investigated referred to the
standard operating point of the reduction phase with a
blowing rate of V_ Ar = 120 m3/minstp (blowing rate
100 pct) and a ll level of Hst = 2.096 m (ll level
100 pct). In melt shop practice, the torque was measured
with the help of strain gauges (DMS) arranged on the
torque support arm.
Figure 3 shows the water model at the Institute for
Industrial Furnaces and Heat Engineering of the
RWTH Aachen University.[13] The model consisted of
40-mm-thick acrylic glass components (Figure 3(a)).
The vessel was installed in an aluminum frame and
connected to the latter with a tilting axis and a trunnion
ring. The torque support arm used was a square solidmaterial bar (Figure 3(d)). The lower part of the vessel
contained the nozzles, which were adjustable in all
spatial directions (Figure 3(e)). The ow rate of every
nozzle was controlled by a oat-type ow meter. The
whole vessel could be rotated around the trunnion axis.
Nozzle 4 additionally had a pressure sensor and a
thermocouple. With the help of two ultrasonic sensors
arranged above the water surface, the oscillation was
measured with a resolution of 10 Hz and 0.35 mm
(Figure(b)). Four strain gauges were mounted on the
shafttwo were located opposite each otherwhich
detected the shaft load after the signal conversion of the
vertical force Fy and the horizontal force Fz. Another
pair of strain gauges measured the force Fts acting on the
torque support arm (i.e., its deection). To increase the
resolution of the strain gauges, the trunnion axis was
designed as a hollow shaft. Capacitative sensors to
measure the acceleration and an inductive position
transmitter were arranged at the bottom of the vessel.
The model was equipped with an ink-injection system
for residence time distribution (RTD) investigations.
A dened amount of ink was injected, and the variation
in brightness was detected at a certain position in the
water domain. For ow visualization, an argon-ion
VOLUME 41B, APRIL 2010399

Table I.

Nomenclature of the 120-ton AOD Converter and the 1:4 Scale Water Model

Scale
Inner volume
Liquid steel volume (100 pct)
Liquid steel mass
Vessel tilting angle
Vessel height
Vessel cone diameter
Vessel inner diameter
Vessel bottom diameter
Vessel outer diameter
Liquid steel level (100 pct)
Distance between bottom and trunnion axis
Number of side-wall nozzles
Nozzle height above vessel bottom
Inner diameter of the side-wall nozzle
Outer diameter of the side-wall nozzle
Length of the side-wall nozzle
Angle between the side-wall nozzles
Density of melt/water
Density of slag
Density of argon/air at standard conditions
Process gas ow rate (100 pct)
Cooling gas ow rate (100 pct)
Modied Froude number

S
V0
V
M
c
H
D0
D1
D2
D3
Hst,w
Ht
nn
Hn
Dn,in
Dn,out
Ln
an
qst,w
qsl
qAr,air,stp
V_ Ar;air
V_ Ar
Fr

m3
m3
t
deg
m
m
m
m
m
m
m

m
m
m
m
deg
kg/m3
kg/m3
kg/m3
m3/minstp
m3/minstp

Converter

Model

1:1
63.1
17.3
120
08
7.430
1.182
3.724
2.885
5.100
2.096
3.007
7
0.498
0.016
0.018
1.2
18
7033
2990
1.784
120
15.4
3268

1:4
0.846
0.292
0.274
015
1.858
0.300
0.931
0.721
1.011
0.524
0.752
7
0.125
0.002

0.1
18
998

1.293
0.286

3268

design torque support arm, whereas in the water model


test, the standard deviations were used because of their
more accurate statistical statement.
B. Side-Wall Nozzle and Gas Injection
The process gases were seen as the driving force
behind the melt homogenization. In addition to the
bubble size, the length L with which the gas jet (index g)
penetrated the liquid (index l) played an important part.
The behavior of gas bubbles in the melt was characterized by weight, buoyancy, and inertia forces, and
therefore, the following modied Froude number:
Fr
Fig. 2Simplied, rotational symmetric geometry of the 120-ton
AOD converter.

continuous wave laser was used, and the tracers used


were polyamide particles (q = 1000 kg/m3).
Following the AODs standard operating practice
during the reduction phase (Figure 1) and taking into
account the laws of similarity, various blowing rates and
ll levels of 50 pct to 150 pct each and dierent nozzle
congurations were investigated. Because of the dierent positions of the strain gauges, no direct comparison
could be made between the operational and the water
model measurements. In addition, the focus in operation
was on the maximum values of the torque oscillation
because they supplied important reference values for the
400VOLUME 41B, APRIL 2010

qg u2g
;
ql gDn

where ql  qg

6

The operating point is Fr = 3268. In the following


consideration, the shroud tube of the side-wall nozzle
was neglected. The gas (O2, N2, Ar, and/or air) exited
the nozzle as a continuous jet at almost sonic speed
(jetting regime) and, in the uid (melt or water),
disintegrated into a multitude of individual bubbles
(bubbling regime).[14] The bubbles rose as a heterogeneous buoyant plume, which entrained the surrounding
melt. To prepare for the CFD model, the dierent
behavior of the various gases throughout the process
had to be neglected. For example, an oxygen bubble
directly reacts with the chromium contained in the
melt. The exothermal reaction (2) results in local
temperatures of more than 2500 C result, which cause
the gas bubble to expand rapidly. If the bubble enters a
cooler region afterward, then it abruptly will contract
again. This eect is superimposed by the increase in
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 31:4 scale water model of the 120-ton AOD converter equipped with innovative measurement technique; standard conditions of the water
model are V_ air = 0.29 m3/minstp (seven nozzles), Hw = 0.524 m.

bubble size and the decrease in ferrostatic pressure. In


contrast to this nding, an argon bubble used to
intensify mixing behaves like an air bubble rising in
water because it will not react with the melt. As a
general rule, it can be said that the higher the blowing
rate, the smaller the bubbles are. The diameter of gas
bubbles in a melt is between approximately 10 and
100 mm.[14]
The penetration length of the gas jet is important
because the ow structure depends on the spatial
distribution of the induced plume. The plume generates
two asymmetric mixing zonesa smaller zone on the
nozzle wall and a larger zone opposite the nozzle wall.[6]
In this article, the mean penetration length L of the gas
jet into the liquid uid was established according to
Hoefele and Brimacombe[12] because this approach
agreed most with our own experience and is expressed
as follows:
 0:35
qg
0:46
L 10:7Fr Dn
7
ql
As the Fr number, nozzle diameter Dn, and qg/ql
ratio increase, the length of penetration L rises. To
ensure that the ow structure in the model (index m)
and original (index o) converter were similar, the ratio
Lm/Lo must have corresponded to the models scale of
1:4. After rearranging Eq. [7], the nozzle diameter
METALLURGICAL AND MATERIALS TRANSACTIONS B

Dn,m obtained in the water model is expressed as


follows:


qAr qw 0:35
Dn;m 0:25Dn;o
8
qst qair
Based on the use of argon as a process gas and the relevant uid properties, Dn,m = 2 mm. The nozzle length
is selected at Ln,m = 0.1 m. These values ensured that
the penetration lengths of the jet in the model were
similar to the original. The standard volumetric ow
rate to be set at the model nozzle is expressed as
follows:
s


qAr qw Dn;m 5 _ 2
_
Vn;Ar
9
Vn;air
qair qst Dn;o
From the equation of continuity, the mean outlet
speed from the nozzle is expressed as follows:
un;air

4V_ n;air p0 T
pD2n;m p T0

10

with the standard values p0 and T0.


To review the Hoefele and Brimacombe approach,[12]
the penetration of an air jet into water was evaluated
statistically.[13] The theoretical penetration length LAr
of an argon jet in the AOD converter, as ascertained by
VOLUME 41B, APRIL 2010401

Fig. 4Penetration length LAr of argon gas into melt for nozzle
length Ln = 1.2 m and diameter Dn = 16 mm. Penetration length
Lair of air into water for nozzle length Ln = 0.1 m and diameter
Dn = 2 mm.

Eq. [7], was compared with the length Lair of a uid


dynamic similar air jet in the water model (Figure 4).
The operating point of the AOD converter and the
equations for the penetration lengths of argon and air
are plotted. Generally, the penetration length was low.
At the operating point of the AOD process, it was
LAr = 0.39 m for the argonmelt system and
Lair = 0.086 m for the airwater system. The theoretical
values according to Hoefele and Brimacombe agreed
with the measured penetration lengths in the water
model. When raising the nozzle pressure beyond the
design condition, the measured data were leveling, and
Lair remained more or less constant. No penetration of
the air jet beyond the center of the vessel and up to the
opposite wall was observed. The air bubbles always rose
close to the nozzle wall and generated a large-space,
quasistationary vortex. This general behavior coincided
with the real wear pattern of the refractory lining
observed in AOD practice.
In a preliminary study, the injection of argon through
a single side-wall nozzle was calculated with the help of
a transient, two-phase CFD simulation according to the
EulerEuler approach. Here, a complete system of uiddynamic equations was solved for each uid-ow phase.
The injection of a compressible, cold gas, which attained
roughly sonic speed at the nozzle outlet and then entered
a highly viscous incompressible hot-melt domain at
an approximately 7000 times higher density, required
extreme demands on the CFD solver. Thus, a numerical
trick was employed to stabilize the solution and make it
convergent. The results obtained are variables p, T, q, u,
and Ma at the nozzle outlet but, above all, the
penetration length LAr of an argon jet in melt. These
results serve as input data for a simpler and less
compute-bound injection model of Lagrangian formalism in which the gas downstream from the nozzle was
specied in the form of an empirical bubble distribution.
This process is referred to as the discrete phase model
(DPM) (see Section IV).
Figure 5 shows results of the EulerEuler simulation
shortly after the argon injection began into a rectangular
402VOLUME 41B, APRIL 2010

Fig. 5CFD simulation (EulerEuler approach) of argon gas injected into the melt domain, Hst = 2.096 m, Ln = 1.2 m,
Dn = 0.016 m. The ow structure at t = 0.5 s after the start of gas
injection is shown. Argon boundary conditions at the nozzle inlet
are pAr,in = 9.3 bar (above ambience), TAr,in = 20 C.

vessel lled with melt up to an elevation Hst = 2.096 m.


The heat transfer between gas ow and surrounding
refractory material has been considered. Argon gas
exited the nozzle at pAr,ex = 7.6 bar, TAr,ex = 25 C,
qAr,ex = 4.2 kg/m3, and uAr,ex = 447 m/s, the pressure
drop of the nozzle was Dpn = 1.7 bar. The penetration
length varied between LAr = 0.35 and 0.4 m, which
agrees with the values in Figure 4. The gas acceleration
in the nozzle agreed with Fannos theory. Cooling
caused by expansion of the gas in the nozzle was
compensated partly by the wall heat ux from the
refractory lining. Outside the nozzle, an underexpansion
up to Ma = 1.4 can take place, which because of the
high static dierential pressure between the gas at the
nozzle outlet (pAr,ex = 7.6 bar) and the pressure in
the melt (pst,ex = 1.45 bar, above ambience), seemed to
be realistic. This phenomenon also is supported by the
abrupt heating of the argon gas.
In operational practice, a small portion of the melt
solidies at the tip of the nozzle because of cooling
eects. The adhered steel develops into a tubular shape
and extends the nozzle length by a few centimeters.
This nding causes the process gas to penetrate the melt
METALLURGICAL AND MATERIALS TRANSACTIONS B

a little deeper and the exothermal reactions of the


oxygen in front of the nozzle to take place further away
from the refractory lining.

IV.

NUMERICAL MODELS

The CFD simulations are based on the unsteady


Reynolds averaged NavierStokes (URANS) equations
in conjunction with a turbulence model. For the water
model, the realizable ke model was used, and for the real
AOD converter, the shear stress transportscale adaptive simulation (SST-SAS) model was used.[1518] In
subjective terms, the SST-SAS model shows a ner
resolution of the stochastic phenomena than other
turbulence models, especially at the phase boundaries.
For the multiphase ow (waterair and/or meltslag
gas), the volume of uid (VoF) model was used.[19] This
model is a special EulerEuler approach for two or more
immiscible uids in which a surface-tracking technique
is applied to a xed Eulerian mesh to predict the
interface position between the uids. A single set of
momentum equations was shared by the uids, and an
additional equation for the volume fraction of each uid
in each grid cell was tracked throughout the computational domain.
The uid movement, especially near the phase interfaces, induced time-depended forces and torques on the
converter wall. These forces caused the converter
oscillation. In every time step of the CFD simulation,
~s;i
~p;i and the friction force F
the total pressure force F
acting on the wall element and, thus, the resulting force
~i are calculated as follows:
F
~i F
~p;i F
~s;i
F

11

~i around point Pt on the trunnion


The tilting torque M
axis (Figure 2) was calculated as follows:




~i ~
~p;i ~
~s;i
M
r0;i  ~
rt  F
r0;i  ~
rt  F
12
Here ~
r0;i is the position vector of the ith wall element
that referred to the coordinate origin P0. With the help
of a user-dened function (UDF), the center of mass
~s , and M
~
~p ; F
for all phases and the integral variables F
were established from the transient simulation. The
inuence of the oscillating converter on the uid was
neglected (i.e., one-way coupling of domains). In
evaluating the forces and torques that are variable in
terms of time, the standard deviation (rms) was used as
a quantitative criterion because the mean value is, by
denition, equal to zero. However, this nding also
means that peak forces that occurred in real operation
were ltered out.
Because of the computing time, the gas injection
process was not realized by a EulerEuler technique but
by a so-called DPM method. This method provided for
the specication of an empirical bubble distribution
close to the nozzle outlet. The DPM model performed
Lagrangian trajectory calculations for the dispersed
phases (i.e., particles, droplets, or bubbles), which
included coupling with the continuous phase. To this
end, the inertia, mass, buoyancy, and drag forces acting
METALLURGICAL AND MATERIALS TRANSACTIONS B

on the bubble were considered.[15] To include the


inuence of the turbulence on the bath movement, a
stochastic bubble tracking method was used. Interactions between the continuous and the dispersed phase
also were taken into account (two-way turbulence
coupling). Small bubbles generally behaved like rigid
spheres, whereas larger bubbles became cup-shaped with
a constant drag coecient (cd,b). In this article, a
Broder[20] approach was used in which the drag coecient was a function of the bubble Reynolds number.
The expansion of the bubble while rising in the melt also
was taken into account. To this end, the local bubble
diameter Db as a function of the vertical y-coordinate
was described on the simplied assumption of an
isothermal change of state and is expressed as follows:


q gy  Hn 0:333
Db Db;Hn 1 l
13
p
Db,Hn is the local bubble diameter at the level of bubble
injection Hn (Figure 2). For the water model, the
increase in diameter was approximately 1.6 pct, and
for the melt, it was around 35 pct. The gas bubbles were
added to the relevant uid in the form of a Rosin
Rammler distribution. This method described the cumulative mass portion of the gas bubbles with a diameter
larger than Db, as a function of Db, and a parameter that
stated the width of the distribution. For the water
model, bubble diameters of Db,air = 1 to 35 mm with
b;air = 20 mm were assumed, and for the AOD conD
b;Ar = 40 mm. The
verter Db,Ar = 5 to 70 mm with D
bubbles were eliminated from the computing domain at
the phase boundary (i.e., waterair or meltgas and/or
slaggas). The process data and uid properties needed
for the DPM were obtained from the described
preliminary simulation using the EulerEuler approach.
The numerical grid of the water model consisted of
around 0.5 million hexahedron cells and, for the AOD
converter, of around 1.5 million hexahedron cells.

V.

SOME LAYOUT CRITERIA

The AOD process was aected largely by the converter geometry (Table II and Figure 6). Ecient reaction kinetics required a rapid, intensive mixing of the
melt. The mixing time and mixing intensity depend, inter
alia, on the ratio of the bath diameter and the bath

Table II.

Basic Process Conditions that Aect the Flow


Structure in an AOD Converter

Fill level of liquid steel


Homogenization process
Mixing time
Wear rate of the refractory lining
Movement of the free surface
Distance between melt center
of gravity and trunnion axis
Vibration intensity

 st <1
D=H

 st >1
D=H

high
good
short
high
intensive
low

low
bad
long
low
weak
high

high

low

VOLUME 41B, APRIL 2010403

Fig. 6Fundamental ow structure for dierent geometries of the AOD converter.

 st : The results shown in Reference 21


height D=H
demonstrate that the mixing time increased with an
 st ratio.
increasing D=H
 st = 1,
Unaected by geometric restrictions, ideally (D=H
Figure 6(a)) a circular primary vortex with a high
angular momentum and a high kinetic energy results,
which covers a large region of the bath. On the opposite
converter wall, the melt ows downward. The melt
velocity adjacent to the wall is high, mixing is good,
 st rises with a
and the mixing time is low. When D=H
constant melt capacity (Figure 6(b)), the vortex forms
an elliptical shape, and the kinetic energy is dissipated.
The rotational energy decreases and the mixing
 st
time increases. With an even higher ratio of D=H
(Figure 6(c); e.g., at the end of a converter campaign),
the maximum rising velocity of the gas bubbles may no
longer be achieved and part of the kinetic energy of the
plume becomes lost for the melt drive. Because the
 st ratio then is unfavorable to form a circular
D=H
vortex, the low-momentum melt close to the free
surface will not reach the opposite converter wall but
returns to the interior and the primary vortex decays
into several secondary vortices. Inactive dead-ow
regions are induced, which are only slowly involved in
the chemical reactions. Although the melt velocity near
the converter wall is low, it minimizes wear, but mixing
is poor.
Closely associated with transient ow phenomena are
refractory wear and converter vibrations. Reactions
taking place at temperatures higher than 2500 C,
process-related temperature variations, high-melt ow
velocities, and turbulence lead to erosion of the nozzleside converter wall, especially at the meltslaggas phase
boundary. Typical wear rates are 35 mm/heat. Wear of
the nozzle-side converter wall is counteracted by a
thicker refractory lining. High velocities near the wall,
 st ratio, increase the wear
which occurs for a low D=H
404VOLUME 41B, APRIL 2010

rate and vice versa. Moreover, the dissolution capacity


of the slag decreases with an increasing converter
diameter, which may require the excessive addition of
a uxing agent. Consequently, all requirements, a short
mixing time, low wear rates, and vibration cannot be
satised at the same time. For a freshly relined AOD
 st  1.5 to 2 is a good compromise.
converter, D=H
The side-wall nozzles have to be arranged at a low
level to ensure a high rising level (Hst Hn) of gas
bubbles (Figure 6(d)) and to avoid dead ow regions.
The nozzles may not be located at a low level either, or
else the converter bottom will erode. If nozzles are
installed at a high level (Figure 6(e)), then a part of the
ferrostatic height to drive the melt will be lost, and an
inactive ow region will be induced close to the bottom.
Investigations are currently under way that deal with the
concentric arrangement of nozzles in the converter
bottom (Figure 6(f)). A positive inuence is expected
on the reaction kinetics as well as on the vibration
amplitude. The inclination of the refractory wall is
relevant for mixing because the wall is not allowed to
counteract a uniform, concentric vortex formation. The
usual number of degrees is 20 deg to 35 deg. In the
course of the converter campaign, the refractory lining
will adapt itself to the existing ow structure and adopt
a dished contour.

VI.

RESULTS

A. Plant Test
SMS Siemag carries out vibration analyses on converters across the world, with statistical evaluation
based on several converter campaigns. The subject of
the analysis in the following discussion is the 120-ton
AOD converter shown in Figure 2.
METALLURGICAL AND MATERIALS TRANSACTIONS B

To counteract the high wear rate, many plant owners


incline the converter vessel by 5 deg to 10 deg during the
process. The intention is to put the plumes away from
the refractory lining (i.e., to reduce the wear rate above
the side-wall nozzles and to improve the conditions for
the species conversion, or both). Operational measurements made with both a vertical (c = 0 deg) and an
inclined (c = 8 deg) converter clearly indicate the tendency toward a higher vibration intensity when the
vessel is inclined.
Figure 7 shows the dierence DM = Mc = 8 deg
Mc = 0 deg of the measured torque standard deviation
M both with an inclined and a vertical vessel for the
process stages. This result is the mean deviation from
the zero position (i.e., the higher the standard deviation,
the more intensive the vibration). To minimize the
inuence of the progressing wear of the refractory
lining, the tests were made one immediately after the
other after expiry of 40 pct and 60 pct of the converter

campaign. Although the results were not necessarily


consistent at the various process stages, all cases indicate
a signicant increase in vibrations when the converter is
inclined. For example, the standard deviation during the
reduction phase after 60 pct of the converter campaign
rose by DM = 8.4 pct.
The operational tests also prove that the measured
torque (i.e., maximum value as well as standard deviation) at a variation range between 110 t and 145 t is
independent of the heat size. The variation between
8 pct and 21 pct of the nominal melt capacity is
aliated with a ll level variation of DHst = 0.13 m
and DHst = 0.32 m, respectively. This unexpected result
is reduced to the fact that the movement of the free
surface and, thus, the vibration level is barely aected by
the variation of the ll level around the nominal height.
Only at low melt heights will the surface waves be
damped by the converter bottom. This behavior is
conrmed largely by water model tests, which show only
a moderate dependence of the vibration amplitude on
the ll level so that the amplitude decreases slightly with
a rising ll level (see Section VI-B).
Moreover, the operational tests indicate that the
torque oscillation was highest at the start of the
converter campaign, which decreased by around 25 pct
by the end of the campaign. This statement is based on
the evaluation of approximately 450 converter heats and
has been conrmed recently by other plant tests.
Generally, tendency is clear for N2 to induce slightly
higher oscillation amplitudes than Ar. The measured
frequency of the 120-ton AOD converter dominating in
the frequency spectrum is typically at around 2.5 Hz.
B. Water model

Fig. 7Dierence of the torque standard deviation DM = Mc = 8 deg


Mc = 0 deg measured for the upright and the inclined 120-ton AOD
converter after 40 pct and 60 pct of the converter campaign.

Figure 8 shows laser-light-sheet visualizations at a


variable blowing rate (for 100 pct ll level) and a
variable ll level (for 100 pct blowing rate). The air jets
penetrated the water model from the left and quickly

Fig. 8Laser-light-sheet visualization at the water model for (a) dierent blowing rates and (b) ll levels. Standard blowing rate with seven nozzles V_ air = 0.29 m3/minstp (blowing rate 100 pct) and standard ll level Hw = 0.524 m (ll level 100 pct).
METALLURGICAL AND MATERIALS TRANSACTIONS B

VOLUME 41B, APRIL 2010405

collapsed into a variety of individual bubbles ascending


on the vessel wall and caused a primary vortex that
rotated clockwise. The size and location of the vortex
remain nearly unchanged. A rough, heavily bubbling
water surface was generated, which may be regarded as
the cause of the vessel vibration. The higher the blowing
rate, the more intensive the vortex rotation and the more
turbulent the water became. Even at a high air-inlet
pressure into the side-wall nozzle, the air jet did not
penetrate far into the water. At a ll level of 75 pct

(HDw = 2.10), several stochastic individual vortices were
formed, and the primary vortex arbitrarily changed its
sense of rotation. From a ll level of around 100 pct

(HDw = 1.58) on, the primary vortex statistically behaved
almost at a steady state. The higher the ll level, the
faster the primary vortex rotated. Here again, the
movement of the water surface increased but not to
the same extent as with increasing the blowing rate.
Several investigations have been performed on the
water model. For the following statements, the measured data are always referred to the standard (index:
stand) mode of operation at a blowing rate of 100 pct,
ll level of 100 pct, and c = 0 deg. For example, the
relative deviation of the x-force component was
DFx = (Fx Fx,stand)/Fx,stand.
(a) When inclining the water model converter by
c = 15 deg toward the nozzle wall, the rms value relative to the standard mode of operation increased by
around 18 pct (DFy,rms = +6 pct, DFz,rms = +32 pct,
DFts,rms = +16 pct). This order of magnitude was
obtained in the operational test as well (see
Section VI-A). For all blowing rates and ll levels
investigated, the detected model forces and, thus, the
vibration amplitudes were higher with an inclined
vessel rather than with an upright vessel.
(b) When simulating the inuence of the dampening
slag layer, which is highly viscous during the
decarburization phase, in the water model by adding
wooden spheres of a suitable density onto the water
surface, the rms value was reduced by around 10 pct
(DFy,rms = 7 pct, DFz,rms = 9 pct, DFts,rms =
15 pct). Similar results were obtained for other
blowing rates, ll levels, and c values.
(c) When the torque support arm was removed, the
vertical force DFy.rms decreased by around 13 pct,
whereas the horizontal force DFz,rms increased by
around 56 pct. This behavior applied to all blowing
rates, ll levels, and c values.
(d) An additional weight of 50 kg xed to the bottom of
the model supplied DFy,rms = 3 pct, DFz,rms = +14 pct
and DFts,rms = +6 pct.
(e) Figure 9 illustrates that the higher the blowing rate
V_ air becomes, the shorter the mixing time s95 is. As
the ll level Hw increases, the mixing time s95 rst
increases then remains constant for high ll levels.
(f) The natural frequency fE of the water model test rig
was measured for the neutral position at a ll level of
100 pct. The dominating peak in the frequency
spectrum was fE1 = 4.3 Hz. More sub maxima
were found at fE2 = 5.6 Hz, fE3 = 25.3 Hz, and

406VOLUME 41B, APRIL 2010

Fig. 9Mean mixing time s95 and corresponding standard deviation


measured at the water model for dierent blowing rates V_ air and ll
levels Hw.

fE4 = 29.5 Hz. The test rig was designed with the
help of Pro/E, and the theoretical natural frequency
was calculated from the total stiness and the total
moment of inertia by taking into account the torsional stiness of the trunnion axis and torque support arm. fE,theo = 4.5 Hz agreed with the rst
maximum of the measurement.
(g) For standard conditions, the maximum levels in the
frequency spectrum were found at f1 = 1.0 Hz,
f2 = 4.8 Hz, f3 = 13.5 Hz and f4 = 20.5 Hz, with
dominating f2 and f3. An increase in the ll level did
not reveal a uniform frequency response. Although
the rst maximum f1 increased slightly here, the
values for f2-4 decreased. In the water model, a
variable blowing rate had no inuence on the frequency spectrum.[11]
The results of the vibration and frequency measurement with variable blowing rates and ll levels are
presented in the following sections, which include a
comparison with the results of the CFD simulation.
C. CFD Simulation and Comparison with Plant Test
and Water Model Investigation
The movement of the free surface and the deviation of
the velocity across time for the standard case (blowing
rate 100 pct, ll level 100 pct) are shown in Figure 10.
The centerline of the converter and the ll level in the
rest position also are shown. The plumesof which only
the rear ones are showndo not reveal any individual
bubble but an iso-volume that enclosed the respective
plume with a constant bubble concentration. The colors
of the iso-volume corresponded to the local melt
velocity. In the CFD simulation, the physical boundary
conditions for the RosinRammler distribution of the
injected air bubbles have been adapted so that the
calculated length of penetration Lair agreed with
Lair = 0.086 m as measured on the water model. The

METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 11CFD simulation for the water model with V_ air =


0.435 m3/minstp (blowing rate 150 pct) and Hw = 0.524 m (ll level
100 pct). (a) Distribution of the force Fy according to Eq. [11] and
(b) distribution of the torque My according to Eq. [12]. Minimum
and maximum value of the standard deviation as well as the
calculated frequency f1 = 1.2 Hz and f2 = 2.4 Hz.

Fig. 10CFD simulation for the water model including the phase
interface movement for V_ air = 0.29 m3/minstp (blowing rate 100 pct),
free surface level.
Hw = 0.524 m (ll level 100 pct), and

bubbles rose near the nozzle wall and induced the


clockwise rotating primary vortex described earlier as
well as a small elliptical secondary vortex close to the
wall. The size and location of the vortices that aected
the homogenization process were similar for all blowing
rates and ll levels. When the plumes reached the free
surface, high wave amplitudes with a low-frequency
oscillation were induced. The wall-bounded water surface moved up and down. This behavior corresponded
to the type-B oscillation according to Fabritius et al.[11]
(Figure 10(c)). The type-A oscillation, which means the
swashing back and forth between the vessel walls,
neither was calculated by CFD nor observed in any
water model test. However, the characteristic bubbling
METALLURGICAL AND MATERIALS TRANSACTIONS B

known from the water model cannot be simulated


numerically because the bubbles deed a spatial and
time-dependent resolution. If the free surface above the
plumes rose, then velocities in an upward direction were
generated in the water and vice versa. The higher the
blowing rate, the higher the amplitudes of the surface
are. At a blowing rate of 150 pct, the maximum
deviation from the ll level at rest position on the
side opposite the nozzle wall was, for example,
Dy  20 mm. Adjacent to the wall, high velocities
led to high-wall shear stresses and, in reality, to a high
degree of wear. High shear stresses resulted around the
nozzles as well and in the area of the free surface,
especially on the side of the nozzle wall. The maximum
shear stress was found when the water surface passed
the location of the rest position, because at this time, the
induced velocity reached a maximum. As a result of the
periodic excitation, the system was supplied continuously with energy (i.e., this system was a forced wideband oscillation).
For the water model, Fig. 11 gives an example of the
simulated distribution of force Fy according to Eq. [11]
and the distribution of torque My according to Eq. [12],
VOLUME 41B, APRIL 2010407

based on a blowing rate of 150 pct and 100 pct ll level.


In sum, the simulated forces and torques are at a low
level. The corresponding table shows that maxima result
that are clearly higher than the rms value plotted as a
dash-dot line. This behavior is known from the real AOD
process as well. All rms values of the forces are of the
same magnitude. Oscillation has no preferred direction,
but the wide-band excitation induced a stochastic oscillation with a tumbling movement in all spatial directions.
This phenomenon was reected by the rms values of the
torques. Torques Mx,rms and Mz,rms around the horizontal converter centerlines were generally higher than
torque My,rms around the vertical centerline; the latter
was characterized by Fx and Fz. In most cases, Mx,rms
yielded the highest values, presumably because some
surface waves proceeded preferably from the nozzle side
toward the opposite vessel wall. The frequency spectrum
calculated from the forces and torques showed two
dominating frequencies at f1 = 1.2 Hz and f2 = 2.4 Hz.
Figure 12 shows the comparison between the measured and the calculated forces/torques for the water
model at dierent blowing rates and ll levels. On the
trunnion axis of the model, the components Fy and Fz
have been measured (Figure 2). These values are plotted
in the diagrams of the force distribution as colored
symbols without a connecting line. The detected data
supplied by the strain gauge pair installed on the torque
support arm were too small, and the torque support arm
should be required to be designed as a hollow section in
future investigations. Despite the basically dierent
approach in establishing the occurring forces (i.e., on

the one hand, the direct strain gauge measurement on


the trunnion axis of the physical model, and on the other
hand, the integration of the numerically calculated
pressure and shear stresses), one arrives to the same
~rms . The
order of magnitude and a similar behavior for F
~rms values with a
general tendency (i.e., increasing F
~rms
rising blowing rate V_ air and slightly decreasing F
values with a rising ll level Hw, and starting at a ll
level of around 75 pct) is conrmed by both simulation
methods.
As the blowing rate increases, the forces and torques
and, therefore, vibrations increase because the movement
of the surface becomes more intensive; this behavior is
conrmed by the physical simulation at the water model.
As the ll level increases, the amount of water rises,
thereby reducing the distance between the models center
of gravity and the trunnion axis. Consequently, the
whole system receives more mass inertia. In the physical
model, the increasing movement of the water surface
with a rising ll level was at least partially compensated
by the simultaneously increasing mass inertia so that the
forces, torques, and vibrations decreased slightly. In
regard to the surface movement, the numerical model
revealed a contrary behavior. With small ll levels, an
intensive surface movement could be recognized, which
at approximately 100-pct ll level reached its maximum,
will cause it to drop again while the ll level continued to
~rms
~rms and M
rise. This behavior is reected both by the F
values as well as by the peak forces.
Figure 13 shows the results of the transient, threephase, nonisothermal CFD simulations for the 120-ton

Fig. 12Comparison between experiment and CFD simulation for the water model. In all cases, the basis for the shown data is the standard
deviation (rms value) of the forces and the torques for dierent (a) blowing rates V_ air and (b) ll levels Hw.
408VOLUME 41B, APRIL 2010

METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 13CFD simulation for the original 120-ton AOD converter. The movement of the phase interface and the velocity distribution during the
reduction phase is shown. All forces and torques are determined according to Eq. [14] and are referred to the standard process with
melt,
slag.
V_ Ar = 120 m3/minstp (blowing rate 100 pct), Hst = 2.096 m (ll level 100 pct), c = 0 deg, and lsl = 0.001 Pas;

AOD converter lled with melt (yellow) and a 0.3 m


thick slag layer (red). The simulations were based on the
SST-SAS turbulence model in conjunction with the VoF
and the DPM models. The blowing rate was constant in
METALLURGICAL AND MATERIALS TRANSACTIONS B

all cases; V_ Ar 120 m3/minstp (blowing rate 100 pct). At


the start of the reduction stage (Figure 1) the slag
consists of around 40 pct CaO, 5 pct MgO, 5 pct SiO2,
25 pct Cr2O3, 10 pct FeO, 5 pct MnO, and 10 pct
VOLUME 41B, APRIL 2010409

metallic granules and is highly viscous. The arbitrarily


chosen viscosity of lsl = 50 Pas is that of thick honey,
lsl = 0.05 Pas is like glycerin, and lsl = 0.001 Pas is like
water. For the analyses, no individual components were
used at this stage anymore but rather the integral forces
and torques. The latter were referred to standard
operation at V_ Ar = 120 m3/minstp (blowing rate
100 pct), Hst = 2.096 m (ll level 100 pct), c = 0 deg,
and lsl = 0.001 Pas. The following equation then
applies, for example, to the relative change in the force:

~

F
~stand



F
~

D
F

F
~stand

14


q
2
2
2
~
F
with
F
F
F
rms;x

rms;y

rms;z

Figure 13(a) shows CFD results of the fundamental


inuence of the slag layer being molten during the
reduction phase. As the viscosity of the slag decreased,
the dampening eect of the slag layer declined. Movements at the meltslag and slaggas interface increased, as
did the melt turbulence in the uid domain. At
lsl = 50 Pas the resulting force was around 60 pct, and
the resulting torque was 57 pct lower than the standard
case with lsl = 0.001 Pas. These values are high, yet what
must be considered here is that the numerical model
interpreted the sole uid movement as oscillation without
the dampening inuence of the additional converter mass.
In terms of operating practice, this result means that the
converter vibrationdespite a constant inert gas ow
ratecan only increase because of slag liquefaction,
which recently has been conrmed by operational tests.
Yet, the reduction of the slag viscosity below a certain
level, here lsl  0.01 Pas, was not accompanied by
another increase in vibrations. In sum, the large-space
ow structure was similar to that of the water model.
However, the structure of the large-space primary vortex,
which is characteristic for water, has been lost for melt in
favor of many small individual vortices. The local
turbulence inside the melt domain can be extremely high.
Figure 13(b) shows sequence-vs-time images of the
melt ow with Dt = 0.5 seconds. At some points in time,
the lighter weight slag layer was displaced by the rising
melt and intensive melt splashing may have occurred. A
long-stretched secondary vortex rotating anticlockwise
formed at the nozzle wall. An irregular type-B movement was simulated on the surface.
Figure 13(c) shows the situation for a converter that
has been inclined by c = 8 deg. In this case, the rising
argon bubbles moved away from the nozzle wall. This
movement intensies the bath one more time because
the bubble plumes now have more space available for
moving back and forth. The mean deviation of the freemelt surface around the rest position is about twice as
high as the vertically positioned converter. When
compared with the standard case, (right-hand picture
in

Figure 13(a)) the acting forces
increase by

~
113 pct and torques by D
M
~
122 pct. The
D
F
same applies to the vibration behavior because it
correlates with the forces and torques. This phenomenonpredicted based on CFD simulationswas

410VOLUME 41B, APRIL 2010

conrmed by the operational test in Figure 7. The


torque standard deviation DM, as measured at the
torque support arm, was a mere 8.4 pct. However,
the following two aspects must be considered: On one
hand the measurement was based on the evaluation of
just a few heats and also may vary considerably because
of permanently changing boundary conditions in the
melt shop. On the other hand, the mass inertia of the
converter plant was not taken into account in the CFD
simulation.
Figure 13(d) shows the situation at Hst = 2.620 m (ll
level 125 pct). Although the intensity of the surface
waves subjectively agreed with the standard ll level
and torques
Hst = 2.096 m
(ll

level 100 pct), the

forces

~
28 pct and D
M
~
47 pct when
decreased by D
F
compared with standard operation. It is obviously not
only the pure movement near the
free

surface

of

the melt
~
and D
M
~
, but also
responsible for the amount of D
F
the local turbulence within the melt bath. For the
100-pct ll level, the ow was characterized by a
multitude of turbulent individual vortices, whereas with
the 125-pct ll level it was characterized by a relatively
stable vortex rotating clockwise. Following the consid st = 1
erations from Figure 6, the primary vortex at D=H
 st 1 it became more
was stable, whereas for D=H
unstable. For this reason, the ow structure in standard
 st =
operation (Hst = 2.096 m, ll level 100 pct D=H
1.58, right-hand picture in Fig. 13(a)) was also signicantly more stochastic than the higher ll level
 st = 1.26, see
(Hst = 2.620 m, ll level 125 pct D=H
Figure 13(d)). But then again, higher static forces
resulted in the latter case because of the higher converter
weight.
The melt movement correlates with the minimum and
maximum forces as well as with the rms forces.
Figure 14(a) illustrates the inuence of a variable tilting
angle c by referencing the horizontal component Fy in
the main vibration direction. A process time of 50 seconds was shown after the full transient response of the
converter. Both the mean amplitude and the rms value
clearly increased. This was particularly obvious when
looking at torque Mx, which formed from components
Fy and Fz and increased from 15.9 kN to 59.9 kN
(Figure 14(b)). The frequency established from CFD is
shown in the diagrams as well. For c = 0 deg,
f1 = 0.48 Hz, and f2 = 1.30 Hz, for c = 8 deg,
f1 = 0.47 Hz, and f2 = 0.93 Hz. In all cases, the fast
Fourier transform (FFT) supplied several peaks in the
frequency spectrum. However, it did not matter whether
the FFT was applied to the force, the torque, or the local
melt speed; the result was always the same. Similar
statements applied when the ll level was changed from
Hst = 2.096 m (ll level 100 pct) to Hst = 2.620 m (ll
level 125 pct) in Figure 15. The ll level, however, had a
smaller inuence on Fy,rms and Mx,rms than the tilting
angle.
Table III summarizes the most important results and
compares the resulting minimum and maximum forces
and torques as well as the rms values for the cases
investigated (i.e., the standard case (1), the inclined
converter (2), and the higher ll level (3)).

METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 14CFD simulation for the upright (c = 0 deg) and the


inclined (c = 8 deg) 120-ton AOD converter with V_ Ar =
120 m3/minstp (blowing rate 100 pct) and Hw = 2.096 m (ll level
100 pct). Time-dependent distribution of the (a) forces and (b) torques as well as simulated frequencies.

Fig. 15CFD simulation for


and Hst = 2.620 m (125 pct)
V_ Ar = 120 m3/minstp (blowing
dependent distribution of the
simulated frequencies.

the ll level Hst = 2.096 m (100 pct)


of the 120-ton AOD converter at
rate 100 pct) and c = 0 deg. Time(a) forces and (b) torques as well as

Table III. Results of the CFD Simulation for the 120-ton AOD Converter with Dierent Process Conditions
Component
Min./max. peak
rms value
Min./max. peak
rms value
Min./max. peak
rms value

Fx [kN]

Fy [kN]

Fz [kN]

Mx [kNm]

My [kNm]

Mz [kNm]

Case

76/81
35
126/127
68
38/28
13

1460/1379
19
1450/1388
24
1860/1744
18

34/38
13
89/292
50
46/67
20

38/51
16
359/124
60
52/64
21

0.04/0.10
0.04
0.14/0.18
0.07
0.09/0.05
0.03

96/97
43
152/154
82
38/27
13

1. Fill level 100 pct


c = 0 deg
2. Fill level 100 pct
c = 8 deg
3. Fill level 125 pct
c = 0 deg

VII.

CONCLUSIONS

By introducing process gases under the surface of the


melt bath, the AOD process causes considerable vibrations that may be intensive enough to make components
of the structural steelwork, foundations, and plant
components suer from them. For safe design, it is
important to understand the causes and relationships of
these vibrations. For this reason, the chemical-physical
processes in the AOD converter have been investigated
by a combination of operational measurements, physical
simulation on a downscaled water model, and numerical
simulation using ANSYS FLUENT. In particular the
CFD simulation provided an eective tool to make the
METALLURGICAL AND MATERIALS TRANSACTIONS B

AOD process more transparent. Some inconsistencies


should be noted between the results of the dierent
simulation approaches. This concerns the height of the
measured and calculated torques for the 120-ton AOD
converter as well as the respective frequencies. Future
investigations to clarify these discrepancies are currently
on their way.
The present work allows the following important
conclusions for the AOD process:
(a) The converter oscillation is essentially caused by
transient movements of the phase-boundary surfaces. The process gases injected through the
submerged side-wall nozzles and the plumes,
VOLUME 41B, APRIL 2010411

respectively, act as a vibration stimulus with wideband excitation. Characteristic torque oscillations
are highly determined by the dened process standard (i.e., by the oxygen and inert-gas (N2, Ar) ow
rate at the blowing stages and their allocation to
dened bath-carbon contents). The vibration
amplitude was highest during the last stage of the
dynamic blow.
(b) The length with which the inert gas jet penetrates the
melt domain can be determined with good approximation by using an empirical approach according
to Hoefele and Brimacombe.[12] Even at a high inlet
pressure, the penetration length of a highly pressurized inert gas is not more than approximately 0.4
to 0.5 m.
(c) The vibration amplitudecharacterized by the
maximum value on one hand, and by the rms value
on the other handdepends on the process gas type
and ow rate. Vibrations are low when the proportion of oxygen in the process gas is high. Vibrations
become more intensive as the amount of inert gas in
the process gas rises, and the inert gas ow rate
increases. The type of inert gas has an inuence on
the oscillation. N2 has a tendency to induce slightly
higher oscillation amplitudes than Ar. The detected
frequency is independent of the blowing rate.
(d) The operational tests demonstrate that the maximum level of the torque oscillation does not depend
on the heat size when the latter varies between
8 pct and +21 pct of the nominal heat size. Both
the physical and the numerical simulation indicate
that the standard-deviation forces and torques drop
slightly with a rising ll level. The relating frequency
spectrum, though, does not reveal a clear-cut
behavior. However, the frequency is dropping
slightly with rising ll level.
(e) A small proportion of the vessel vibrations is caused
by the oxygen injected through the side-wall nozzles
and its reaction with carbon and a clearly larger
proportion by the injection of inert gas.
(f) The slight inclination of the converter vessel during
the ongoing process as practiced by many steelmakers
to put plumes away from the refractory lining
increases converter vibrations. This general phenomenon, whose occurrence is independent of the blowing
rate and ll level, indeed may intensify homogenization caused by the higher melt turbulence.
(g) Torque oscillations are highest with a freshly relined
converter, which decrease with increasing wear of
the refractory lining. During the converter campaign, the oscillations decrease by around 25 pct.
(h) The CFD simulation shows, among others, that
even at a constant inert gas ow rate, the oscillation
of the converter vessel can increase because of the
slag liquefaction.
(i) The higher the blowing rate, the shorter the mixing
time. As the ll level rises, the mixing time rst
increases and remains almost constant for high ll
levels.

412VOLUME 41B, APRIL 2010

TABLE OF SYMBOLS
~
r0
V_
a
q
c
s95
a
c
cd
D
f
f
fE
F
Fp
Fr
Fs
g
H
K
L
M
n
n
p
P0
pCO
S
T
u
V
x,y,z

Position vector
Volumetric ow rate
Angle between the side-wall nozzles
Density
Tilting angle of the converter vessel
Mixing time
Activity
Mass concentration
Drag coecient
Diameter
Coecient of activity
Frequency
Natural frequency (eigenfrequency)
Force
Pressure force
Modied Froude number
Friction force
Acceleration due to gravity
Height
Equilibrium constant
Length
Torque
Molar mass
Number of Nozzles
Pressure
Origin of coordinates
Partial pressure
Scale
Temperature
Velocity
Volume
Coordinates
INDICES

0
air
Ar
b
C
CO
ex
g
i
in
l
m
n
o
st
stand
theo
ts
w
x,y,z

Mean value
Referred to standard uid ow conditions
Air
Argon
Bubble
Carbon
Carbon monoxide
Exit
Gas
Index
Inlet
Liquid
Model
Nozzle
Original
Liquid steel
Referred to standard operation
According to theory
Torque support arm
Water
Components in direction of coordinates

METALLURGICAL AND MATERIALS TRANSACTIONS B

REFERENCES
1. H.W. Grenfell, D.J. Bowen, and C. Mcqueen: Proc. Natl. Open
Hearth Basic Oxygen Steel Conf., 1977, pp. 20921.
2. F. Mucciardi, V. Krujelskis, and E. Palumbo: Proc. 5th Process
Technology Conf., Detroit, MI, 1985, pp. 8590.
3. M. Ohnishi, J. Nagai, T. Yamamoto, K. Nakai, H. Tacke, and
R. Tachibana: Iron Steelmaker, 1983, vol. 10 (8), pp. 2834.
4. M. Bjurstrom, A. Tilliander, M. Iguchi, and P. Jonsson: ISIJ Int.,
2006, vol. 46 (4), pp. 52329.
5. T.M.J. Fabritius, P.A. Kupari, and J.J. Harkki: Scand. J. Metal.,
2001, vol. 30, pp. 5764.
6. T.M.J. Fabritius, P.T. Kurkinen, P.T. Mure, and J.J. Harkki: Iron
Steelmaker, 2005, vol. 32 (2), pp. 11319.
7. T.M.J. Fabritius, P.T. Mure, P.A. Kupari, V.A. Juntunen, and
J. Juhani: Steel Res. Int., 2001, vol. 72 (7), pp. 23744.
8. A. Tillander, P.G. Jonsson, T.L.I. Jonsson, and S. Lille: Iron
Steelmaker, 2002, vol. 29 (2), pp. 5157.
9. A. Tillander, T.L.I. Jonsson, and P.G. Jonsson: ISIJ Int., 2004,
vol. 44 (2), pp. 32633.
10. J.H. Wei, H.L. Zhu, S.L. Yan, X.C. Wang, J.C. Ma, G.M. Shi,
Q.Y. Jiang, H.B. Chi, L.B. Che, and K. Zhang: Steel Res. Int.,
2005, vol. 76 (5), pp. 36271.
11. T.M.J. Fabritius, P.T. Mure, and J.J. Harkki: ISIJ Int., 2003,
vol. 43 (8), pp. 117784.

METALLURGICAL AND MATERIALS TRANSACTIONS B

12. E.O. Hoefele and J.K. Brimacombe: Metall. Trans B, 1979,


vol. 10B, pp. 63148.
13. H. Pfeifer, A. Ruckert, C. Wuppermann, and M. Schnitzler:
Report 08/03, Institute for Industrial Furnaces and Heat
Engineering, Aachen, Germany, 2008.
14. B. Deo and R. Boom: Fundamentals of Steelmaking Metallurgy,
1st ed., Prentice Hall, Upper Saddle River, NJ, 1993.
15. ANSYS FLUENT: Users Guide, ANSYS FLUENT, Ann Arbor,
MI, 2008.
16. L. Davidson: ECCOMAS CFD Conf., Egmond aan Zee, The
Netherlands, 2006, pp. 120.
17. J.H. Ferziger and M. Peric: Computational Methods for Fluid
Dynamics, 2nd ed., Springer-Verlag, New York, NY, 1999.
18. F.R. Menter and Y. Egorov: Aerospace CFD Conf., Paris, France,
2007, pp. 117.
19. C.W. Hirt and B.D. Nichols: J. Comp. Phys., 1981, vol. 39,
pp. 20125.
20. D. Broder: PhD Dissertation, Martin Luther Universitat, HalleWittenberg, Germany, 2003.
21. H.J. Odenthal, W.H. Emling, J. Kempken, and J. Schluter:
Association for Iron & Steel Technology Conf., Indianapolis, IN,
2007, pp. 117.

VOLUME 41B, APRIL 2010413

You might also like