An Improved Lance Design For Hot Metal De-Sulphurisation: Ironmaking & Steelmaking
An Improved Lance Design For Hot Metal De-Sulphurisation: Ironmaking & Steelmaking
To cite this article: P. Tripathi, D. S. Kumar, R. Sah & V. R. Sekhar (2016): An improved
lance design for hot metal de-sulphurisation, Ironmaking & Steelmaking, DOI:
10.1080/03019233.2016.1214379
Article views: 14
Download by: [University of California, San Diego] Date: 20 September 2016, At: 12:40
An improved lance design for hot metal
de-sulphurisation
P. Tripathi, D. S. Kumar∗ , R. Sah and V. R. Sekhar
Hot metal de-sulphurisation is a dip-lance process involving the pneumatic injection of fine-grained
de-sulphurisation reagents into the molten metal. For maximum efficiency the particles must be
dispersed in the ladle as widely as possible to increase the total interfacial area which is
primarily controlled by the lance design. Seven different lance configurations were modelled and
simulated to determine the most efficient design using physical and mathematical modelling
approach. A 0.25 scale plexi-glass model of the 100 T hot metal ladle was fabricated for the
study. Residence time and mixing time studies were carried out using the electrical conductivity
measurement technique through stimulus response of injected saturated salt solution.
Mathematical modelling approach using momentum balance was used to simulate fluid flow
profile of lance-ladle assembly under operating conditions using computational fluid dynamics
package ANSYS-CFX. Based on the studies a new curved port lance has been designed which
resulted in uniform and swirling flow profile inside the ladle without rotating the lance. Injection
through the new lance increased the residence time of the particles and reduced the dead
zones. The new design was fabricated and experimented at de-sulphurisation stations and has
resulted in reduced flux consumption and treatment time.
Keywords: De-sulphurisation lance, Physical modelling, Mathematical modelling, Ladle
of hot metal and steel by injection of powdered agents in through fabrication of complex nozzle designs in physical
the literature.1–5 In this process, dispersion of fine particles model and CFD analysis. The aim was to improve the
into the molten metal bath increasing the interfacial area mixing time and residence of the particle in the melt
is greatly responsible for the efficiency of the processes. through an optimum port design. The optimum design
These carbide particles are poorly wetted by molten was subsequently validated through plant trials.
metal and their densities are very low compared to that
of the molten metal which restricts its uniform distri-
bution. It has been reported4,5 that stirring helps in uni- Lance port design configurations
form mixing of reagents in the ladle and better Based on the literature review and practical experiences at
desulphurisation kinetics is achieved. A recent develop- the de-sulphurisation unit, seven different types of con-
ment is the use of rotating lances6,7 for increased turbu- figurations were selected for study as listed below and
lence, but has maintenance and cost issues. Majority of shown in Fig. 1. The central tube diameter and the port
steel plants still continue to use static lances. In static dimensions were kept constant for all the configurations.
lance injection process lance port design governs the injec- Configurations (a) S lance and (b) T lance are the most
tion profile of the desulphurisation reagents into the ladle. commonly used designs in the steel industry. Configur-
The ability of the static lance to produce better mixing has ation (c) and configuration (d) are inspired from BOF
to be given due importance in the design, when compared lance designs. Configuration (e) is an extension of con-
to improving the other parameters impacting de-sulphur- figuration (b). Configuration (f) and configuration (g)
isation such as higher carbide quality, slag conditions, are completely new designs experimented for the first
temperature and mechanical agitation which involves time.
higher cost and yield losses. (a) 1 vertical straight port (S lance)
A large number of studies8–10 have been performed on (b) 2 straight ports in mutually opposite directions (T
the steelmaking processes and the role of oxygen lance lance)
and its design through physical and mathematical model- (c) 2 ports inclined at 45° angle in mutually opposite
ling, in particular, over the years but the de-sulphurisation directions (T-45 lance)
lance is less investigated. Many physical and mathemat- (d) 2 ports inclined at 15° angle in mutually opposite
ical [computational fluid dynamics (CFD)] studies11–13 directions (T-15 lance)
for static lance gas injection systems focused on gas– (e) 4 ports in mutually perpendicular directions (TT
metal dual-phase interactions in ladle and bubble beha- lance)
viours, but the effect of lance port configurations is rarely (f) 2 curved ports in mutually opposite directions (C-45
addressed. Additionally, the published research result lance)
does not suggest any optimised design for external de-sul- (g) 2 spiral ports in mutually opposite directions (H-45
phurisation and more relevant experimental data is lance)
necessary to verify the published CFD model simulations.
In the present work, effect of different lance port designs
on improving the de-sulphurisation efficiency has been
investigated by means of physical and mathematical Water modelling experiments
(CFD) modelling techniques. It may be the first time A 0.25 scale plexi-glass model of the 100 ton hot metal
when all major possible designs have been investigated ladle was fabricated for the study. The details of actual
ladle and model vessel are given in Table 1. The air supply
line to the lance assembly consisted of a pressure regulator
and air flow rotameter. A flow metre capable of control-
ling flow rates between 0 and 50 lpm was used for flow
control. The lance tip was thread fitted to the central
lance to carry out experiments with different lance tip 3 Water model measurements
designs. Conductivity measurements were carried out
using the electrical conductivity measurement technique circuit time for each design was noted and plotted for
through stimulus response of injected saturated salt sol- further studies. The mixing process in a lance-ladle injec-
ution. Flow visualisation studies were carried out using tion system consists of two components – mixing inside
coloured KMnO4 solution. the plume volume and outside it. There is a significant
The vessel was filled with water up to the height that difference between the mixing times at both the regions
represents the level of the metal bath. The position of and hence the measurement in the slowest outside region
the lance and water level were fixed for the entire set of is critical for estimating the bulk mixing time. Through
experiments. The extent of mixing was determined by some series of experiments it was found that the conduc-
conductivity measurements using saturated solution of tivity probe dipped 800 mm into the ladle at 50 mm away
NaCl as tracer and a standard conductivity metre from the wall of the vessel shown as position ‘B’ in Fig. 3
(Cyber Scan CON 110) for measuring the conductance was the most suitable location as it took relatively higher
of the solution. Data generated was captured using a com- time for complete homogenisation compared to other
puter equipped with data acquisition card and dedicated locations. 50 ml tracer was used in each experiment and
software. The experimental setup is shown in Fig. 2. Com- conductivity of the solution was recorded continuously
pressed air was introduced through the top of the lance at thereafter till the reading becomes constant. Lance tips
a predetermined rate of 15.5 lpm using the flow metre and were changed and the same process was repeated for all
was monitored continuously during the experiments. the configurations. The mixing time for each set of exper-
When steady state was attained, the tracer was injected. imental conditions was calculated from the conductivity
The short circuit time (T1) and internal mixing time vs time plot. The results of physical modelling were com-
(T2) were measured and compared for all the configur- pared with CFD model predictions.
ations. Short circuit time is a direct indicator of amount
of un-reacted carbide going directly into the slag, whereas
internal mixing time indicated the uniformity of the car- Mathematical model
bide mixing. For determining the short circuit time the
tracer was injected through the lance and the conductivity Mathematical modelling approach was used to simulate
probe was positioned at the centre of lance and vessel wall fluid flow profile of lance-ladle assembly under operating
just below the water surface shown as position ‘A’ in conditions and compared with the water modelling exper-
Fig. 3. After tracer injection, first major change in con- iments. Commercially available CFD code ANSYS-CFX
ductivity was recorded by conductivity metre. Short 14.0 was used for simulations. In this study, a 3-D CFD
model was developed to analyse the flow field in the liquid
iron bath inside a ladle with different nozzle design con-
figurations. A full scale mathematical model was devel-
oped using momentum balance considering hot metal as
a Newtonian and incompressible fluid.
Governing equations
The governing equations of fluid flow employed for devel-
opment of CFD model are listed below.
∂r
+ ∇ · (rU) = 0 (1)
2 Experimental setup ∂t
where, r is fluid density, t is time and U is the flow velocity For simulating interfacial mass transfer between the
vector field. continuous and dispersed phases, the Particle model was
In the case of incompressible flows ρ is a constant, used. The specific surface area was calculated by Equation
hence, the above mass continuity equation simplifies to 8 assuming that dispersed phase is present as spherical
a volume continuity equation. particles of mean diameter db in the continuous phase.
∇·u=0 (2) 6rb
Aab = (8)
db
The momentum equation
∂(rU)
+ ∇ · (rU · U) = −∇p + ∇ · t + SM (3)
∂t Drag model
where, p is the static pressure, t is the shear stress and SM Particle model was used to model inter-phase drag on dis-
is the momentum source term. The stress tensor, t, is persed fluid phase by selecting a drag coefficient of 0.44.
described as For spherical solid particles that are sufficiently large
for inertial effects to dominate viscous effects (the inertial
2
t = m ∇U + (∇U)T − d∇ · U (4) or Newton’s regime), the drag coefficient becomes inde-
3 pendent of Reynolds number; whereas in the transitional
where, m is the dynamic viscosity. region between the viscous and inertial regimes, both vis-
cous and inertial effects are important. Hence, the drag
coefficient is a complex function of Reynolds number
Buoyancy
and is described by Schiller Naumann drag model as:
For simulating the effect of buoyancy due to difference in
densities of the interacting fluids, a source term is added 24
CD = (1 + 0.15 R0.687
e ) (9)
to the momentum equations as follows: Re
SM,buoy = (r − rref )g (5)
The density difference (r − rref ) was evaluated using
Full Buoyancy model for modelling multi-phase flow.
Modelling and boundary conditions
Turbulence model The model was developed in full scale. No symmetry
planes were considered so as to allow both symmetrical
Due to presence of re-circulatory zones with high Rey-
as well as non-symmetrical flow conditions depending
nolds number in the continuous fluid domain, k − 1
on the selected turbulent conditions. The present scope
model was used to model turbulence as it offers a good
of study includes comparison of fluid flow profile, hence
compromise in terms of accuracy and robustness. k is
isothermal and steady state conditions were assumed for
the turbulence kinetic energy and is defined as the var-
all the simulations. It allows for the assumptions that
iance of the fluctuations in velocity.
the density of molten steel does not change appreciably
∂(rk) m during steady state casting conditions. The simulation
+ ∇ · (rUk) = ∇ · m + t ∇k + Pk involved a multi-phase study with following three phases,
∂t sk
phase 1 – molten hot metal: continuous phase; phase 2 –
+ Pkb − r1 (6) nitrogen (carrier gas): dispersed gaseous phase; and phase
1 denotes the rate at which the velocity fluctuations dissi- 3 – solid calcium carbide particles: dispersed solid phase.
pate. Non-homogeneous model for multi-phase flow was used.
It enables different velocity fields for individual phases.
∂(r1) m k − 1 turbulence model with high turbulence intensity
+ ∇ · (rU1) = ∇ · m + t ∇1
∂t s1 was used for continuous phase and dispersed phase zero
1 equation model was used for dispersed phases.
+ (C11 (Pk + P1b ) − C12 r1) (7) The mass flow rate inlet boundary condition was used
k
for the inlet, and degassing boundary condition was
where, C11 , C12 , sk and s1 are constants. used at the outlet (top of the ladle). This option resulted
in the boundary acting as an outlet for dispersed phases
Multi-phase model and a free slip wall for the continuous phase. As the
The simulation involved a multi-phase study with follow- ladle is open to atmosphere, 1 atm pressure was assumed
ing three phases, phase 1 – molten hot metal, phase 2 – at the outlet boundary. Three-dimensional tetra-prism-
nitrogen (carrier) gas, and phase 3 – solid calcium carbide pyramid elements were used to generate unstructured
particles. Molten hot metal was taken as the continuous mesh for ladle and lance domains. A minimum mesh
liquid phase with nitrogen gas and solid calcium carbide quality of 0.5 was ensured to get higher accuracy. Sub-
particles as the dispersed gaseous and solid phase respect- merged entry nozzle and mould walls were assumed as
ively. As all these three phases have different material no slip walls. N2 gas mean bubble size was assumed to
properties and a distinct flow field, hence, the inhomo- be 25 mm and CaC2 particle mean diameter assumed to
geneous model was used for simulating multi-phase flow be 100 µm. Buoyant conditions were simulated based on
in the ladle. This allows each fluid to possess its own density difference. The density of hot metal was used as
flow field and different fluids interact via inter-phase buoyant reference density to avoid round-off errors during
transfer terms. calculations.
Simulation process
The liquid iron and gas were considered stationary in the
initial condition. Nitrogen gas and solid calcium particles
were injected through the lance ports at specified flow
rates as per the actual plant practice. The momentum
from the impinging jet stirs the molten hot metal in the
ladle and a fully developed flow gets established after
some time. The established state was taken as an initial
guess and then solved again iteratively under steady
state conditions to obtain final flow profile in the ladle.
Port exit velocities and flow profiles were compared for
each lance design configuration. 4 Short circuit time for different ports
Mixing study
Configuration (f) C-45 and configuration (g) H-45 were
Once the flow field in the ladle has reached a quasi-steady
found to have the largest short circuited flow time and is
state, a tracer having the same properties as the molten
expected to give the largest residence time to the carbide
hot metal was introduced in the ladle as an additional
particles in the ladle. Interestingly, the least short circuited
mass input source at a designated location just below
flow time was found for configuration (b) T shaped lance.
the hot metal surface. The tracer was added as a pulse
This can be attributed as the prime reason for higher
input for just one second with zero velocity fields. Due
amount of un-reacted carbide in HMDS slag.
to the flow conditions prevalent in the ladle, it is carried
Mixing time is an indicator of fluid flow profile and dif-
away with the currents and gets mixed in the ladle. The
fusion characteristics inside the vessel. The mixing time
concentration of the tracer is continuously monitored at
should be minimum for faster de-sulphurisation process.
a specified location in the ladle. The time after which
Lower mixing time indicates that de-sulphurised hot
the tracer concentration achieves 95% of the final con-
metal from the vicinity of lance can be quickly diffused
stant value is taken as the mixing time in the ladle. The
towards the dead zones. Figure 5 shows the time taken
CFD solver settings used for simulations are listed in
(in seconds) for uniform mixing under different port con-
Table 2.
figurations. Results show that configuration (e) TT – the
four port lance and configuration (f) C-45 lance have
Results and discussions the lowest mixing time and are expected to have higher
reaction kinetics. Configuration (a) S – the straight hole
The rate of de-sulphurisation reaction is controlled by the lance has shown the maximum mixing time and is
diffusion of sulphur to the surface of CaC2 rather than the expected to have least efficiency.
rate at which the CaC2 and sulphur reacts. For diffusion Based on the short circuited flow studies and mixing
controlled reaction the rate of reaction is affected by the time studies it can be deduced that configuration (f) C-
intensity of stirring. The degree of de-sulphurisation is 45 lance will be an optimised design which will ensure
proportional to the amount of reagent exposed to the maximum utilisation of the injected carbide and will
melt. For the same flow rate the lance design which have best reaction kinetics inside the vessel. The reason
imparts higher stirring over a larger area in the ladle for better performance of configuration (f) is demon-
will have better de-sulphurisation. Stirring in ladle–lance strated in flow visualisation studies. Even at the constant
system is a strong function of the plume dynamics flow rate the plume diameter and the plume velocity var-
which causes the re-circulation. The plume dynamics is ied significantly with the port design. The plume
measured by the plume diameter and the plume velocity. dynamics difference between configuration (f) C-45 –
Higher plume velocity results in reagent loss to the slags swirling effect lance and commonly used configurations
whereas higher plume diameter results in lower re-circula- (a) S – the straight hole lance and (b) T – two port
tion. Water modelling experiments were aimed to calcu- lance and is shown in the Fig. 6. It shows that the injecting
late the short circuit time and internal mixing time for reagents are more dispersed in a configuration (f) C-45
different configurations which can impart higher stirring lance (wider plume with lower velocity) than compared
over a larger area. The short circuit time is the time to the lances being currently used. This C-45 lance
taken for the reagent to reach the top after injected
from the lance port. This time should be maximum for
optimum utilisation of the reagent. Higher short circuited
time indicates higher residence time for the injected
reagent, hence giving more time for complete reduction
and better sulphur removal. Figure 4 shows the short cir-
cuit time for different ports designs experimented.
adjacent to the lance. However, configurations f C-45 and expected to improve the effectiveness of carbide and the
g H-45 lance give a slight swirling action to the hot metal reaction kinetics during hot metal de-sulphurisation pro-
which results in more uniform stirring of melt in the ladle. cess. The cross-sectional comparison of the commonly
This swirling action helps to reduce the dead flow regions used two port lance and the new swirling effect lance is
thus enabling a better mixing in the bottom portion of the shown in Fig. 10. The velocity flow vectors shown in
ladle. The effect is more uniform in the f C-45 lance Fig. 11 indicate that swirling effect is induced from the
among the two. As the reagents are more spread out by bottom of the ladle.
the time it reaches to the surface, the turbulence at the
top surface is minimised. It can be seen from these figures
that the CFD model predicted very similar flow profiles to Comparison to rotating lances
those observed from the physical model. In the selected lance design (C-45 lance), the swirling
Based on the preceding discussion, the new swirling effect is much smaller when compared to rotating lances,
effect lance has been proposed for plant trials which are but the key advantage is that the effect is limited to the
metal in the lower part of the ladle and the top slag layer is configuration (b) and was replaced by configuration (f).
not vigorously agitated. In case of the rotating lance pro- Ten pieces of new lances were fabricated using the new
cess the hot metal needs to be skimmed prior to de-sul- swirling effect lance design as shown in Fig. 12 and put
phurisation in order to remove high SiO2 containing to trials at one of the de-sulphurisation station in JSW
blast furnace slag. In the rotating lance, de- Steel Vijayanagar works.
sulphurisation process results in increased refractory New designed lance was found to be working well with-
wear, higher temperature and metal losses and has large out any severe operational abnormalities under normal
free board requirement. Hence, shifting to such process conditions. As each hot metal ladle has a different combi-
requires significant changes in ladle and mechanical sys- nation of input ‘S’, temperature, filled level, carryover
tems and is viable option for hot metal desulphurisation slag etc., the performance of the new lance was compared
only when the main target is to produce low sulphur with the old lance under similar conditions. The efficiency
steel (<0.001%) and when process times, temperature of de-sulphurisation process can be measured on the
loss and hot metal loss are not an issue.14 Hence for basis of ‘S’ drop per unit carbide consumption and de-
most of the steel plants the static lance process is the sulphurisation rate constant at the same de-sulphurisation
most effective and flexible process with regard to its ability station under similar conditions with use of two different
to use different reagents, operating methods and process lances. Table 2 shows the comparison of ‘S’ drop per unit
parameters. The proposed lance design will maximise carbide consumption for three different hot metal sulphur
the efficiency of the static lances economically. having temperature >1360°C. ‘S’ drop/unit carbide con-
sumption has improved with the use of new lance design
and is higher at higher input hot metal sulphur. Since
Plant trials the de-sulphurisation reaction is a first order reaction,
The industrial process is more complex and has other change in sulphur concentration is function of time and
additional variables which affect directly or indirectly the rate of de-sulphurisation slows down as the reaction
the de-sulphurisation efficiency such as bath depth, progresses. The rate constant is given by Equation (10)
temperature, slag carryover, and lance vibrations etc. It where k is the rate constant, t is the treatment time and
is therefore important to validate the results at the Sinitial and Sfinal are the initial and final sulphur concen-
plant scale under the operating conditions which helps trations, respectively.
in analysing the limitations regarding the direct appli- %Sfinal
cation in practice. JSW steel was using a T shaped log = −0.434 kt (10)
%Sinitial
lance at its de-sulphurisation units, shown as
12 Fabricated lance
Hot metal ‘S’ range T lance New lance T lance New lance