You are on page 1of 9

Mixing Time and Correlation for Ladles Stirred with Dual

Porous Plugs
JAYANTA MANDAL, SUJOY PATIL, M. MADAN, and DIPAK MAZUMDAR

Bulk mixing times up to a degree of 95 pct were measured in three different, cylindrical-shaped water
model ladles (D  0.60 m, 0.45 m, and 0.30 m, respectively) in which, water was agitated by air intro-
duced through two tuyeres/nozzles placed diametrically opposite at the base of the vessels at 1/2 R
positions. To this end, the electrical conductivity measurement technique was applied. A range of gas
flow rates and liquid depths were investigated (viz. 0.7  L/D  1.2 and 0.002  m (watt/kg) 0.01)
and these were so chosen to conform to the practical ladle refining conditions. In the beginning, extensive
experimental trials were carried out to assess the reliability of the measurement technique. In addition,
some experiments were carried out to determine the location of the probe in the vessel such that measured
mixing times could be interpreted as the bulk mixing times.
It was observed that for smaller gas flow rates (or specific energy input rates), 95 pct bulk mixing
times tend to decrease appreciably with increasing gas flow rates (e.g., mix  Q0.58). However, for
relatively higher flow rates, the dependence was found to be less pronounced, mixing times decreasing
nearly in proportion to a third power of gas flow rates. Similarly, it was found that there exists a critical
gas flow rate for any given vessel beyond which mixing times in dual plug stirred configuration are
somewhat shorter than those in equivalent axi-symmetrical systems. A dimensional analysis followed
by multiple regression of the experimental data (for m  0.07 W/kg) indicated that mixing times in
ladles fitted with dual plugs located diametrically opposite at R/2 locations could be reasonably
described via mix, 95 pct  15Q0.38L0.56R2.0 in which L is the depth of liquid (m), R is the vessel radius
(m), and Q is the ambient flow rate (referenced to mean height and temperature of the liquid). Finally,
the adequacy and appropriateness of the correlation was demonstrated with reference to the experimental
data derived from a 0.20 scale, tapered cylindrical-shaped water model of a 140 T industrial ladle as
well as scaling equations and modeling criteria reported in the literature.

I. INTRODUCTION mixing, as well as to promote better slag/metal intermixing


and to avoid explosive degassing effects under vacuum. Given
THE efficiency of many chemical processing operations such, the relevance of the present study to industrial steel-
carried out in the present day steelmaking ladles is intricately making practice is readily apparent.
related to mixing phenomena. Mixing enhances chemical Joo and Guthrie[2] were among the first to investigate mixing
reactions by bringing in reactants together and removing phenomena in a ladle fitted with twin porous plugs both exper-
products from reaction sites. In addition, it also influences imentally and computationally. Their study indicated that shorter
the extent of thermal and particulate inhomogeneities within mixing times (relative to conventional axisymmetrical/
the ladle. It is therefore desirable to ascertain the extent of asymmetrical gas bubbling) can be achieved by injecting gases
mixing in order to evaluate the process performance of argon- through two porous plugs, located diametrically opposite at
or nitrogen-stirred ladles. midbath radius position (i.e., at 1/2R). Reconstructed from
Mixing phenomena, by and large, have been investigated Reference 2, this is shown in Figure 1, in which mixing times
in aqueous models of gas-stirred ladle systems in which a for an axisymmetrically placed single porous plug/nozzle is
centrally or an asymmetrically placed single porous plug/nozzle shown vis á vis the same for twin plug bubbling as a function
is used to stir the contents of the ladle. A great deal of infor- of net gas flow rates. There, it is at once evident that the twin
mation on these is already available in the literature and sum- plug configuration ensures relatively faster mixing only at mod-
marized by Mazumdar and Guthrie[1] in a review. While fluid erately higher gas flow rates. The figure also appears to suggest
model studies on gas-stirred systems using a single, axisym- that such advantages associated with twin plug bubbling are
metric/asymmetric plug/nozzle have been relatively common, likely to fade away at lower operating gas flow rates. This
not much information on ladles fitted with dual plugs/nozzles aspect, however, was not given sufficient attention by Joo and
is available in the literature. It is to be noted that occasionally Guthrie in their work and has not since been confirmed through
it might be necessary to bubble an industrial ladle with two any subsequent investigation.
or more porous plugs, in order to achieve gentle but rapid In a later work, Zhu et al.[3] carried out similar investi-
gations on ladles fitted with one or more porous plugs. Their
observations reproduced from Reference 3 and shown in
JAYANTA MANDAL, Engineer, is with Ispat Industries, Dolvi, Figure 2 clearly indicate that mixing times, regardless of the
Maharastra, 402107, India. SUJOY PATIL, Graduate Student, M. MADAN, magnitude of gas flows, are smaller in the dual-plug-stirred
Senior Project Associate, and DIPAK MAZUMDAR, Professor, are with
the Department of Materials and Metallurgical Engineering, Indian Insti-
system than those in the equivalent axisymmetrical system.
tute of Technology, Kanpur, 208016, India. Contact e-mail: dipak@iitk.ac.in Such observations are grossly similar to those reported by
Manuscript submitted September 18, 2003. Joo and Guthrie, while some dissimilarity between the two

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 36B, AUGUST 2005—479


above that a physically plausible, effective correlation for
estimating mixing times does not yet exist for such systems.
As such, the influence of the operating variables such as gas
flow rates (Q), depth of liquid (L), and radius of the vessel
(R) are likely to exert on mixing is not known with any
certainty for such a system. Consequently, the primary objec-
tive of the present work has been to experimentally inves-
tigate mixing phenomena in ladles fitted with dual porous
plugs/nozzles over a wide range of operating conditions and,
thereby, to develop an effective correlation for estimating
mixing times in such systems. Toward these, experimental
observations and a discussion on 95 pct mixing times derived
from three different size water model ladles, in which water
is agitated by air introduced through a pair of nozzles located
diametrically opposite at midbath radius position, are pre-
sented in the following Sections III(B) and III(C).
Fig. 1—Experimentally measured 95 pct mixing times in a water model ladle
as a function of gas flow rates for two different nozzle/plug configurations
(reconstructed from Ref 2). II. EXPERIMENTAL WORK
As pointed out already, mixing times were measured in
three different cylindrical vessels (i.d.  0.60 m, 0.48 m, and
0.30 m), in which water was agitated by injecting air or N2
through a pair of nozzles located at the bottom of the vessel
at the midbath radius position. No attempt has been made to
investigate mixing phenomena as a function of nozzle locations
since it is rather well known[2] that porous plugs/nozzles placed
diametrically opposite at midbath radius position ensure the
best mixing conditions. Prior to monitoring mixing, air/N2
was bubbled into the water bath at the desired flow rate for
a few minutes to ensure the stability of the flow in the vessel
as well as to remove any inhomogeneities present in the bath.
The gas flow rates investigated were so chosen to ensure a
gentle stirring condition as is typically encountered in actual
ladle metallurgy steelmaking operations[1] (viz. 0.001 to
0.015 Nm3/t min). It is to be noted here that while lower end
flow rates are typically used in industry for bath homoge-
nization (i.e., of thermal or material), relatively large gas flow
Fig. 2—Experimentally measured 95 pct mixing times in a water model ladle rates are used to achieve enhanced slag-metal reactions. The
(D  0.40 m) as a function of gas flow rates for two different nozzle/plug
configurations.[3] objective of this study was to primarily investigate bath homog-
enization, it was deliberately decided to use smaller flow rates,
embodying a fraction of the entire range of gas flow rates
sets of experimental results,[2,3] particularly at relatively small mentioned previously.
operating flow rates, is evident. Zhu and co-workers also A conductivity probe supplied with a digital conductivity
proposed an empirical correlation for estimating 95 pct mix- meter (Eutech make, Cyberscan 200, Eutech Instruments Pte
ing times in ladles operated with multiple porous plugs/noz- Ltd., Singapore) was employed to record changes in the local
zles (up to a maximum of three plugs) according to ion concentration of a pulse tracer (NaCl or H2SO4) added
directly to a point on the liquid-free surface, lying midway
tmix  8.52m0.33 N 0.33 [1] between the “eyes” of the two surfacing plumes (e.g., the axis
of symmetry). The change in local ion concentration around
Equation [1] suggests that at an equivalent specific poten- the probe tip was measured through the changes in the water’s
tial energy input rate, m, any increase in the number of electrical conductivity and recorded manually via the digital
porous plugs/nozzles is likely to lead to a corresponding conductivity meter. Thus, data on conductivity were collected
increase in mixing times. This is in direct opposition to the typically every 2/3 seconds and simultaneously stored in a
experimental results presented in Figures 1 and 2. It is instruc- computer using the Cyber Comm Portable software, Eutech
tive to note here that Eq. [1] was developed by Zhu and co- Instruments Pte Ltd., Singapore. On addition of tracer, con-
workers in light of empirical correlations proposed several siderable oscillations in the conductivity value were observed,
years back by Nakanishi et al.[4] for their bottom blowing which essentially resulted from the periodic fluctuations of the
steelmaking converter, fitted with multiple tuyeres/nozzles. amount of tracer passing through the probe tip. For each exper-
The brief review presented above evidently suggests that iment, the recording of tracer response was carried out until
more systematic study of mixing phenomena in dual-plug- the concentration of the tracer in the bath was considered to
stirred ladle is required. Similarly, it is apparent from the have reached a homogeneously mixed value. Experimental

480—VOLUME 36B, AUGUST 2005 METALLURGICAL AND MATERIALS TRANSACTIONS B


Fig. 4—Adequacy of the present experimental technique illustrating the
Fig. 3—A typical tracer response curve and the procedure adopted to esti-
correspondence between measured mixing times (for an axisymmetric gas
mate 95 pct mixing times.
injection configuration) and those estimated on the basis of an equivalent
correlation reported by Mazumdar and Guthrie.[5]
data were subsequently analyzed through MICROSOFT
EXCEL* software and a plot between conductivity vs time was Table I. Ninety-Five Percent Mixing Times as Registered
by a Conductivity Probe Placed at Different Locations in the
*MICROSOFT EXCEL is a trademark of Microsoft Corporation, Bath (D  0.60 m, L/D  1.0, and Q  2.83
104 m3/s)
Redmond, WA.
Height of the
generated. This is shown in Figure 3 from which corresponding Probe Tip from the Average
mixing times were estimated. These are defined in the present Bottom of the Probe Mixing
context as the time required for the monitoring point concen- Vessel (m) Location Times (s)
tration to fall continuously within a 5 pct deviation of the well 0.04 A 30
mixed/homogeneous value. By keeping the probe tip immersed B 27
in the slowly moving/mixing region of the vessel (discussed C 23
in Section III), measured mixing times were interpreted as 0.46 A 23
the bulk mixing times. The present approach therefore ensured B 22
that by the time the monitoring point reached a degree of 95 pct C 30
homogeneity, the bulk of the liquid was practically well mixed. Nozzles/plugs are located in 0.53 A 19
this central vertical plane B 15
A set of at least five measurements was made for each exper- C 20
imental condition and thereby an average mixing time was
determined. The variation in successive measurements was
found to be sensitive to gas flow rates as well as vessel size. evidently indicates the reliability of the present approach.
Maximum variation between successive measurements for the It is important to mention here that the tracer addition as
largest vessel (D  0.60 m) was found to be around 15 pct. well as the monitoring point locations in these experiments
(e.g., Figure 4) were essentially identical to those for which
the correlation[5] was originally developed.
III. RESULTS AND DISCUSSION
A. Adequacy of the Experimental Technique B. Determination of the Monitoring Point Location
Since the digital conductivity meter–probe assembly was for Registering Bulk Mixing Times
applied to measure mixing times in the gas-stirred water The bulk mixing times (mixing times registered from a
model ladles for the first time and mixing was monitored probe located in the slowest moving region in the bath) rather
through manual recording of conductivity data (as opposed than local mixing times are a parameter of practical impor-
to continuous measurements, typically used in such studies), tance. However, none of the two previous studies[2,3] specif-
it was decided to assess the reliability of the present technique ically shed any light on the slowly moving/mixing region
prior to any detailed measurement of mixing times. To this in the dual-plug-stirred ladle system. Consequently, flow
end, experiments were carried out in one of the vessels visualization studies (through addition of KMnO4 solution
(D  0.60 m) for a central/axisymmetric gas injection con- to the bath) were carried out to identify the regions that are
figuration and 95 pct bulk mixing times were measured as relatively slowly moving/mixing in the system. Visual obser-
a function of gas flow rates (varying between 1.95 and 3.0
vations indicated that the recirculating fluid moves rather
104 m3/s) and vessel aspect ratio (in the range 0.8 to 1.2). slowly in the vicinity of the ladle bottom, and there, three
In Figure 4, measured 95 pct bulk mixing times for var- distinct pockets of slowly moving regions (e.g., marked by
ious operating conditions are compared directly with equiv- “A”, “B”, and “C” in Table I) appear to exist.
alent estimates derived from a well-established empirical To confirm this and hence to identify the slowest mixing
correlation[5] (viz. mix,95 pct  25.4 Q0.33 L1.0 R2.33). There, region in the vessel, mixing times were measured in the largest
reasonable agreement between measurements and predictions vessel (D  0.60 m and Q  2.83
104 m3/s), keeping

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 36B, AUGUST 2005—481


the probe immersed at nine different locations in the bath. system. Thus, Mazumdar and Guthrie[5] pointed out that in
In all these experiments, the tracer was added through the axisymmetrical gas-stirred systems at low operating gas flow
free surface immediately above the line of symmetry. It should rates, a relatively weak, secondary recirculation typically
be mentioned here that this tracer addition location remained exists at the bottom of the vessel adjacent to the vertical side
invariant throughout the present investigation. In Table I a wall. This disappears beyond a critical gas flow rate leading
summary of these measurements is presented. There, it is to the characteristic flow patterns of an axisymmetrical gas-
evident that mixing is more rapid in regions close to the stirred ladle system (e.g., large recirculating vortices on either
free surface, whereas toward the base of the vessel, mixing side of the line of symmetry).[5] Tracer dispersion tests carried
is relatively sluggish. Furthermore, Table I indicates that out in the larger model ladle indicated that similar phenomena
the registered mixing time is the longest provided the probe’s are at work in the dual-plug-stirred system as well. Conse-
tip is collinear with the plugs and placed near the junction quently, the shift in mixing time vs gas flow rate relationship
of the vertical side wall and base of the vessel. The exact beyond a flow rate of 12 L/min as depicted by Figure 5
dimensionless probe location in the study was r/R  0.89 appears to be the result of a changing hydrodynamic condition
and z/L  0.93. In all experiments, the conductivity probe within the system. Since the exponent on Q beyond 12 L/min
was therefore placed at location A, as shown in Table I. or 2
104 m3/s is practically equivalent to 1/3,[1,6] it is
therefore reasonable to consider the flow phenomena beyond
C. Variation of Mixing Times with Gas Flow Rates this critical flow rate to be essentially dominated by the inertial
and gravitational forces, as discussed in Section III(D).
A series of experiments were carried out in which 95 pct
bulk mixing times were measured in the largest vessel as a func-
tion of gas flow rates. Since mixing time vs gas flow rate rela- D. Development of a Correlation for Mixing Times
tionships are known to be specific[5] to the operating flow in Ladles Stirred with Dual Plugs/nozzles
regimes, a rather wide range of gas flow rates was applied. In Because the mixing time vs gas flow rates relationship in
Figure 5, a log-log plot between 95 pct bulk mixing times and the dual-plug-stirred system is shown to be specific to the oper-
gas flow rates is presented. There, in the spirit of numerous pre- ating flow regimes, it is unlikely that a single empirical cor-
vious studies, two distinct line segments have been fitted through relation could effectively describe mixing times in such systems
the experimental data points. These indicate that initially mix- across the entire range of gas flow rates. Because industrial
ing time decreases rather sharply (e.g., mix  Q0.58) with ladles are massive and kinematic viscosity of steel is small,
increasing net gas flows, up to about 2
104 m3/s (or inertial and gravitational forces in gas-stirred-ladle systems are
12 L/min). Thereafter, the decrease in mixing times with gas likely to be relatively more important (in comparison to vis-
flow appears to be somewhat less pronounced (viz. mix  Q0.35). cous forces). Consequently, from the viewpoint of practical
A similar behavior between gas flow rates and mixing relevance, it is desirable that a correlation of relevance to the
times was also reported by many investigators for axisym- inertial-force-dominated flow regime be developed (m  0.007
metrical gas-stirred-ladle systems.[1] In such context, it has W/kg, as in Figure 5). Toward this, as our first step, a num-
been suggested[6] that a shift in such a relationship essen- ber of experiments were carried out in the remaining two
tially results from a change in the flow pattern within the vessels as a function of gas flow rates to identify the corre-
sponding critical gas flow rates, Qc. As mentioned already, a
set of at least five measurements were made for each experi-
mental condition from which an average mixing time was
determined and considered for subsequent analysis.
In the inertial- and gravitational-force-dominated flow
regime (viz. Froude dominated flows), the thermophysical
properties of the liquid (i.e., in the absence of any upper
buoyant phase) do not influence flow phenomena to any
appreciable extent. Consequently, a functional relationship
between mixing times mix and the key operating parameters,
namely, the gas flow rate, Q, the liquid depth, L, and the
radius of the vessel, R, can be expressed as[7]

tmix  f (Q, L, R, g) [2]

In terms of the relevant dimensionless groups, Eq. [1] can


be conveniently recast in the following form:
tmix2g L a Q2 b
 C0 a b a 5 b [3]
R R gR
in which C0, a, and b are the three constants to be determined
through consideration of experimentally measured mixing
times and the corresponding operating parameters L, R, and
Q, respectively. Thus, on the basis of the experimentally
Fig. 5—Ninety-five percent bulk mixing times in a water model ladle (D  measured 95 pct bulk mixing time and the corresponding
0.60) fitted with twin porous plugs at different gas flow rates. operating parameters, the three dimensionless groups in

482—VOLUME 36B, AUGUST 2005 METALLURGICAL AND MATERIALS TRANSACTIONS B


(a)
Fig. 6—A comparison between experimental 95 pct bulk mixing times in
three different water model ladles with those deduced on the basis of Eq. [5]
illustrating the adequacy of the proposed correlation.

Eq. [3] were estimated and a multiple regression analysis


carried out embodying the Polymath software.[8] On the basis
of the values of C0, a, and b thus derived, Eq. [3] can be
represented in the following explicit form:
tmix,95 pct2 g L 1.12 Q2 0.38
 936 a b a 5b [4]
R R gR
or, alternatively, in terms of the key operating variables, L,
R, and Q as
tmix,95 pct  15Q0.38L0.5R2.0 [5]
The value of the multiple regression correlation coefficient
for Eq. [4] was about 0.70 and the corresponding variance (b)
was 0.009. The adequacy of Eq. [5] with respect to the exper-
imental data derived from the three different water model Fig. 7—Experimentally measured mixing times in the dual-plug-stirred
ladles is illustrated in Figure 6. Included in the same figure ladle (D  0.60 m) as a function of (a) gas flow rates and (b) liquid depths.
are also the upper/lower bound of uncertainty of the corre-
lation (or the fitted line) vis á vis the experimental data Or, alternatively,
points. These indicate that Eqs. [4] and [5] simulate mixing
times within 21 pct and 32 pct. tmix,95 pct  C0Q0.35L0.56R1.93 [7]
To investigate the extent of data scatter in Figure 6, a 0.35
detailed analysis was carried out and the derivation of the To determine the pre-exponent C0, the parameter, Q
correlation examined from a different standpoint embody- L0.56 R1.93, appearing on the right-hand side of Eq. [7] has
ing experimental data derived from only one single vessel been plotted in Figure 8 as a function of 95 pct mixing time,
(i.e., D  0.60 m). Thus, a large number of experiments mix, and a straight line passing through the origin fitted
(over and above those considered in Figure, 6) were car- through the experimental data points. Evidently, C0  mix, 95 pct,
ried out in the larger vessel for various gas flow rates and when Q0.34 L0.56 R1.93  1.0. The C0 was thereby found
liquid depths, and the results thus obtained are shown in to be 12.5 (m0.34 s0.66). The explicit form of Eq. [7] accord-
Figures 7(a) and (b). These appear to suggest that mixing ingly becomes.
times in ladles fitted with dual plug/nozzle follow the rela- tmix,95 pct  12.5 Q0.34 L0.56 R1.93 [8]
tionships mix,95 pct  Q0.35 and mix,95 pct  L0.56, respec-
tively. On the basis of such, one may represent Eq. [3] as It is at once apparent that Eq. [8] has a form practically iden-
tical to that of Eq. [5] and simulates experimental observa-
tmix 2 g L 1.12 Q2 0.35
 C0 a b a 5b [6] tions within 7 pct. In addition to these, a limited number
R R gR of experiments were also carried out in the three vessels

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 36B, AUGUST 2005—483


through regression, a correlation coefficient close to unity
is unlikely to result (viz. Figure 6).
(2) On the other hand, experimental data derived from a single
vessel and subjected to regression is likely to lead to a
significantly higher correlation coefficient (viz. Figure 8).
(3) In essence, however, the predictive capabilities of the
correlation derived from either a single or a set of vessels
are not likely to be markedly different, as mixing models
derived from these two approaches are shown to have
similar functional forms. Indeed, sample calculations
suggest that if estimates of mixing times are to be derived
via Eqs. [5] and [8] over a range of experimental con-
ditions (e.g., vessel aspect ratio and specific gas flow
rates similar to those considered in this work) embodying
a 220-kg water model through to a 220 T industrial
size ladle, these would tend to differ by a maximum of
about 38 pct or so.

Fig. 8—The variation of 95 pct bulk mixing times with the parameter Q0.34
With reference to the experimental data and regression
L0.56 R1.93 for dual-plug gas bubbling (D  0.60 m). results presented so far, it should be noted that flow phe-
nomena in the gas-stirred ladle systems are inherently tran-
sient.[9,10] Thus, long-term and short-term wandering of the
bubble plume, continuous fluctuation of spout, etc. tend to
make the associated flow field transient and, accordingly,
to influence any measurements. The situation is aggravated
further when two gas bubblers, instead of one, are applied.
Of equal importance is the trajectory of the added tracer
(these influence mixing), which was found to be vastly dis-
similar in different size vessels. For example, in the bigger
vessel, the tracer (saturated solution of NaCl) added imme-
diately over the cenline penetrated to about one-quarter of
the bath depth to be entrained by the rising plumes. In con-
trast, in the smallest ladle (D  0.30 m), the tracer was
found to penetrate right up to the bottom of the vessel from
where it dispersed within the system. Furthermore, tracer
trajectories were not reproducible enough and were influ-
enced significantly due to the wandering of the bubble
plumes. These, as one would normally anticipate, are likely
to introduce some error in measurements and uncertainty
in the final conclusions. Given such, the extent of deviation
of the fitted line from the experimental data points can be
attributed to the inherent characteristics of gas-stirred ladle
systems and the experimental procedure. It is instructive to
note here that previous attempts[11] to unify experimental
Fig. 9—Regression equation and its adequacy to experimental mixing times data from a different group of investigators and different
derived from three different water models fitted with dual porous plugs vessel sizes via a single correlation was also shown to be
located diametrically opposite at R/3 positions.
associated with pronounced data scatter, comparable to those
with tuyeres/nozzles located at the R/3 locations. The resul- shown in Figure 6. In contrast, regression of experimental
tant regression equation (correlation coefficient 0.60 and data derived from a single vessel would typically produce
variance 0.01) and its adequacy to experimental data is less scatter, having an uncertainty of the order of 20 pct
illustrated in Figure 9. There, it is readily seen that the or less.[11]
correlation for R/3 location of the plugs also has a form Although mixing time correlations similar to Eq. [5] are
analogous to those of Eqs. [5] and [8], respectively. On the available for axisymmetrical bubble-stirred ladle systems, no
basis of the evidence presented so far, it is therefore reason- such correlation has yet been reported for ladles fitted with
able to conclude that despite moderate scatter of experimental dual plug/nozzle. Therefore, some comparison can at best be
data with respect to the fitted line (viz. Figure 6), the values made between reported correlation on the axisymmetrical gas-
of the pre-exponent, C0, as well as the exponents a and b, stirred system and the present one. To this end, it is interest-
as suggested via Eq. [4], are plausible and physically realistic. ing to note some similarity between the present correlation
Given such, it is reasonable to assume that and the one (viz. mix,95 pct  25.4Q0.33 L1.0 R2.33) developed
earlier by Mazumdar and Guthrie[7] for the axisymmetrical
(1) If experimental data on mixing times from different size gas-stirred ladle system (e.g., ladles fitted with a central
vessels are applied to derive an empirical equation tuyere/nozzle). It is readily apparent that the values of the

484—VOLUME 36B, AUGUST 2005 METALLURGICAL AND MATERIALS TRANSACTIONS B


pre-exponent (25.4 vs 15) and the exponent on gas flow rate in the dual-plug-stirred system at low operating gas flow, a
(0.33 vs 0.38) in the previous and present correlation are significant portion of the added tracer (added through the
vary similar. Furthermore, a near 1/3 exponent on Q in surface over the line of symmetry) tends to get caught
Eq. [5] appears to confirm that the operating flow regimes in between the two rising plumes. Furthermore, the transport
the dual-plug-stirred systems is dominated by the inertial of the tracer to the probe tip gets delayed, owing to a rela-
and gravitational forces. In contrast, the exponent on the depth tively small angular velocity prevalent in the system at such
of liquid, L, in Eq. [5] is seen to be substantially smaller (by flows. In contrast to this, at equivalent gas flows, the tracer
about a factor of 2 or so). Clearly, the advantage afforded by added over the eye of the plume in an axisymmetrical gas-
taller vessels during dual-plug bubbling seems to be some- stirred bath is transported to the probe tip relatively easily
what limited, in comparison to an equivalent axisymmetrical first via the strong radial surface flows and, subsequently,
system, as is exemplified by a relatively smaller exponent via the downwardly moving fluid adjacent to the vertical
on the depth of liquid, L, in Eq. [5]. This is to be expected side walls. As a consequence of such, the registered bulk
since plume-plume and plume-wall interactions become more mixing times in the axisymmetrical system particularly at
intense with increasing liquid depth (note that the average low gas flows can be expected to be relatively smaller. How-
diameter of plume increases with increasing liquid depth in ever, as the gas flow rate is increased, a strong three-dimen-
the vessel). sional flow pattern (which is characteristically absent in
axisymmetrical gas-stirred systems) sets in the dual-plug-
stirred system, facilitating ready transport of the tracer to
E. A Comparison of the Present Work with the
the probe tip from all regions within the vessel. This latter
Axisymmetrical Gas-Stirred Ladle System
phenomenon appears to be responsible for the observed
A series of experiments were also carried out under con- shorter mixing times in the dual-plug-stirred system at
ditions identical to those in Figure 5 by injecting gas into the relatively high gas flow rates.
system through a centrally fitted nozzle. In Figure 10, a com- The correlation proposed in the present study, the experi-
parison between 95 pct bulk mixing times for axisymmetrical mental data presented so far, together with many previous
and dual-plug gas bubbling is shown for a wide range of gas studies reported in the literature[1] clearly indicate that dimen-
flow rates. There, mixing times in the axi-symmetrical system sions of the ladle (viz. L and R) together with the operating
are somewhat shorter for small gas flow rates than those of gas flow rates exert considerable influence on mixing in
the dual-plug gas injection configuration. However, as the gas ladles. However, such explicit influence of the vessels’
flow is increased, such differences tend to fade away, and dimensions on mixing has not been accounted for in the
eventually mixing time in the dual-plug-stirred system becomes empirical equation of Zhu et al.[3] mentioned earlier. Simi-
shorter. It is instructive to note from Figure 10 that the gas larly, while the present correlation indicates the beneficial
flow rate beyond which dual-plug bubbling ensures superior effects of the dual plug on mixing, particularly in the inertial-
bath mixing is nearly equivalent to the critical flow rate sug- force-dominated flow regime, the correlation of Zhu et al.
gested in Figure 5 (e.g., 12 L/min). As a consequence, it is (viz. Eq. [1]) indicates quite the opposite. It is reiterated that
therefore reasonable to assume that the beneficial effects (to Eq. [1], in contrast to the results presented so far, suggests
the tune of 20 to 25 pct) of dual-plug gas bubbling on mixing that increasing the number of plugs for the same net gas
are limited to the inertial and gravitational force dominated flow rate is not beneficial for exacerbating mixing in the
flow regime alone (m  0.007 W/kg) gas-stirred ladle system.
To rationalize such observations, dye tracer (KMnO4) dis-
persion tests were carried out for the two different nozzle F. An Assessment of the Present Correlation and Its
configurations at various gas flow rates. It was observed that Applicability to Tapered, Cylindrical Vessels
In addition to the arguments and analysis presented so far
in favor of the correlation developed in this study, it is also
important to demonstrate the internal consistency and ade-
quacy (Eq. [5]) from a purely theoretical standpoint. Thus,
applying Eq. [5] between the model and full scale system,
the ratio between the corresponding 95 pct bulk mixing times
can be expressed as
Qmod 0.38 Lmod 0.5 Rmod 2.0
a b a b a b
tmix,mod
[9]
tmix, f s Q f s L f s R f s

L mod Rmod
Incorporating   l together with the scaling
L f s Rf s
equation Qmod  2.5 Qf s in Eq. [9], it can be readily shown
that the ratio between 95 pct bulk mixing times in the model
ladle and the full scale system can be correlated explicitly
via the geometrical scale factor, , according to
tmix,mod
Fig. 10—Variation of mixing times with gas flow rate for two different  l0.5 [10]
gas injection configurations (i.e., axisymmetrical vs dual). tmix, f # s

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 36B, AUGUST 2005—485


This latter equation, as one would note here, can be readily cessing units tend to make direct experimental observations
derived from the first principles considering geometrical and difficult, if not impossible. Naturally therefore comprehensive
dynamical similarities between the model and the full scale experimental investigations encompassing high-temperature
systems.[12] industrial data have been few and far between. Nonetheless,
Finally, mixing times in a 140 T industrial ladle fitted with limited evidence available in the literature[1,13] does appear
dual plugs were also investigated experimentally via a 0.2 to suggest that results derived from appropriately scaled down
scale water model ladle. The operating parameters of the water model studies provide useful quantitative information
model vis á vis those of the full-scale systems are summarized of the actual steelmaking operations. A detailed discussion
in Table II. There, the operating gas flow rates in the model on the subject has been recently published in a review by
ladle were estimated in accordance with the scaling rela- Mazumdar and Evans.[13] It is however sufficient to note that
tionship:[12] Qm  2.5 Qf s. Moreover, the industrial as well water modeling in conjunction with computational fluid
as the model vessels, as reflected from Table II, are slightly dynamics (CFD) is now routinely applied in industrial R&D
tapered, since tapered rather than perfectly cylindrical shapes and academia alike to investigate, optimize, and design steel-
are more typical of industrial practice. Parallel to the exper- making processes. To this end, it should be emphasized here
imental measurements, equivalent estimates were also derived that the correlation developed in this study (and in fact, prac-
in a straightforward fashion via Eq. [5] embodying a mean tically all the mixing time correlation reported in the literature)
vessel radius, Reff. Results thus obtained are shown in Table III. is truly applicable to an idealized, slag-free situation. This
There close agreement between experimental measurements seriously impairs the applicability of such correlations to
and prediction is readily evident. This further suggests that industrial ladles, since slag rather than no slag is more typ-
Eq. [5] is equally effective in tapered, cylindrical-shaped ical of industrial practice and the presence of slag is known
ladles. With reference to the experimental results presented to considerably influence mixing in ladles. It is therefore
in Table III, it is important to mention here that the minimum apparent that in order to increase the generality of the pre-
experimental mixing corresponding to Q  2.8
104 m3/s sent mixing time correlation and many others reported in the
and L  0.7 m appears to be the result of experimental error literature, the influence of the upper slag phase on mixing
or uncertainty, as mixing time has been shown to decrease must be realistically accounted for in such correlation. To
with increasing gas flow rates. date, such a correlation has not been possible. Efforts are
As a concluding point, visual opacity, relatively large size, currently ongoing in our laboratory to develop correlations
and high operating temperatures prevalent in liquid steel pro- for fluid mixing in ladle covered with an upper slag phase.

Table II. Physical Dimensions of the Full Scale (140 T) and


Model (0.2 Scale) Ladle Systems IV. CONCLUSIONS
0.20 Scale Full Scale An experimental investigation on mixing phenomena in
Parameters Model System a ladles fitted with dual plugs has been carried out. From
Base diameter, m 0.52 2.68 the present study, the following conclusions can be drawn.
Top diameter, m 0.58 2.99
Tapering angle, deg 2.34 2.34 1. Mixing time depends on the gas flow rate and vessel size.
Liquid depth, m 0.62 3.20 For a given vessel geometry, mixing time depends on gas
Dimensionless porous plug 0.58* 0.50 flow rate according to mix  Q0.58 for low operating flow
location from the axis rates. However, beyond a critical gas flow rate (m 
of the vessel 0.007 W/kg), the relationship changes to mix  Q0.35.
Gas flow rates, m3/s 1.3
104 8.05
103 2. Beyond the critical gas flow rate, an invariant flow pattern
(corrected to mean height to to in the system is established. The critical flow rate
and temperature of the liquid) 4.2
104 25.14
103 corresponds to the onset of the inertial-and gravitational
*The porous plugs located at the midbath radius position of a perfectly force-dominated flow regime.
cylindrical-shaped ladle corresponded to 0.58R position for the present mar- 3. In the inertial- and gravitational force-dominated flow
ginally tapered vessel. The effect of such small distortion was ignored.
regime, 95 pct bulk mixing times in ladles fitted with

Table III. Experimentally Measured 95 Percent Bulk Mixing Times for the Tapered Model Ladle (Table II)
and Their Comparison with Those Predicted from the Present Correlation (mix  15Q0.38L0.5Reff2.0)

Net Gas Flow Experimental 95 Pct Bulk


Rate (m3/s) Liquid Depth (m) Effective Radius (m) Mixing Time (s) Predicted Mixing Time (s)
Q L Reff mix [15Q0.38L0.5Reff2.0]
2.00
104 0.60 0.273 34.3 36.7
2.33
104 0.60 0.273 32.3 34.6
2.66
104 0.60 0.273 30.6 33
3.33
104 0.60 0.273 28.0 30
1.887
104 0.707 0.275 33.5 29.5
2.839
104 0.707 0.275 24.1 25
3.006
104 0.707 0.275 25.7 24.7
3.334
104 0.707 0.275 28.2 24

486—VOLUME 36B, AUGUST 2005 METALLURGICAL AND MATERIALS TRANSACTIONS B


dual plug/nozzle (placed diametrically opposite at midbath m specific potential energy input rate, W/kg
radius positions) can be reasonably described in terms of  mixing time, s
the three key parameters, L, R, and Q (referenced to mean
height and temperature of the liquid), via mix,95 pct  Subscripts
15Q0.38L0.5R2.0 mod model
4. Mixing times in axisymmetrical gas-stirred systems at f s full scale
relatively high gas flows (m 0.007 W/kg or so) are mix mixing
somewhat longer than those in dual-plug-stirred systems.
The trend is, however, opposite for smaller operating flow REFERENCES
rates.
5. It is shown that mix, 95 pct  15Q0.38L0.5R2.0 can describe 1. D. Mazumdar and R.I.L. Guthrie: Iron Steel Inst. Jpn. Int., 1995, vol. 35,
pp. 1-20.
mixing times in tapered, cylindrical-shaped ladles with 2. S. Joo and R.I.L. Guthrie: Metall. Trans. B, 1992, vol. 21B, pp. 765-78.
equal effectiveness. 3. M. Zhu, T. Inomoto, I. Sawada, and T.C. Hsiao: Iron Steel Inst. Jpn.,
6. The proposed correlation is theoretically consistent and 1995, vol. 35, pp. 472-79.
suggests that for physical and mathematical modeling 4. K. Nakanishi, T. Saito, T. Nozaki, Y. Kato, and K. Suzuki: Proc. Steel-
making Conf., AIME, Warrendale, PA, 1982, pp. 101-08.
of dual-plug-stirred systems Qmod  2.5 Qf s can be used 5. D. Mazumdar and R.I.L. Guthrie: Metall. Trans. B, 1986, vol. 17B,
to scale up/down gas flow rates. pp. 725-33.
6. S. Asai, T. Okamoto, J. He, and I. Muchi: Trans. Iron Steel Inst.
Jpn., 1983, vol. 23, pp. 43-50.
LIST OF SYMBOLS 7. D. Mazumdar and R.I.L. Guthrie: ISS Trans., 1999, vol. 9, pp. 89-96.
8. Polymath software:http://www.polymath-software.com.
C0 dimensionless constant defined by Eq. [3] 9. M. Iguchi, K. Nakamura, and R. Tsujino: Metall. Mater. Trans. B,
1998, vol. 29B, pp. 569-75.
D vessel radius, m 10. D. Mazumdar, C. Seybert, D. Steinfardt, and J.W. Evans: Iron Steel
g acceleration due to gravity, m/s2 Inst. Jpn., 2003, vol. 43, pp. 132-34.
L depth of liquid in the vessel 11. M. Iguchi, T. Kondoh, and K. Nakajima: Metall. Mater. Trans. B,
N number of tuyere or nozzle 1997, vol. 28B, pp. 605-12.
12. D. Mazumdar, H.B. Kim, and R.I.L. Guthrie: Ironmaking and Steel-
Q net gas flow rate, m3/s (referenced to mean height making, 2000, vol. 27, pp. 302-09.
and temperature of the liquid) 13. D. Mazumdar and J.W. Evans: Iron Steel Inst. Jpn. Int., 2004, vol. 44,
geometrical scale factor (Lmod/Lf s) pp. 447-61.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 36B, AUGUST 2005—487

You might also like