Removal of Inclusions Using Micro-bubble Swarms in a Four-strand, Full-scale, Water Model Tundish

Abstract

Water model experiments were performed in a full-scale, delta-shaped water model tundish, in order to study the removal of inclusions by micro-bubbles. Micro-bubbles were generated using a specially designed ladle shroud with twelve laser-drilled orifices. Gas flow rates, injection positions and multi-port injection were all taken into consideration to create different bubble conditions. Bubbles were recorded using a high speed camera and post-processed with commercial software, Image J. Hollow glass borosilicate microspheres, smaller than 100 µm, were used to simulate inclusions, and detected, in-situ, using a new generation of the Aqueous Particle Sensor, APS III. The results revealed that the effect of micro-bubbles on inclusion removal depends greatly on the gas injection protocols used. The optimum gas flow rate was an intermediate value, which indicates a minimum particle number density, n_p, of about 7.85/ml. This results from the counter-balancing effects of bubble sizes against the total number of bubbles. The highest inclusion removal rate was 80%, when gas was injected through the four ports located closest to the slide gate, at a gas flow rate of 0.2 L/min.

1. Introduction

With ever increasing requirements for “clean steel”, the metallurgical functions of a tundish is highly valued. The non-metallic inclusions in liquid steel, including both endogenous and exogenous impurities, can adversely affect physical and chemical properties of the final product. Hence, the efficiency of inclusion removal has been one of the most important criteria in evaluating the metallurgical performance of any given tundish design. Numerous studies have been conducted to optimize the designs of tundishes using various flow control devices,^{1,2,3,4,5,6,7,8,9,10)} such as weirs, dams, impact pads, turbulence inhibitors, etc. In order to promote the floatation of inclusions, these devices can improve the removal of larger inclusions, greater than say 50 μm, but do not necessarily promote the removal of smaller sized inclusions, owing to their extremely low Stokesian velocities.¹¹⁾ H. V. Atkinson and G. Shi¹²⁾ have mentioned the fatigue strength of steel is to a large extent determined by the sizes of inclusions. There is still a limit on the maximum size of inclusions permissible in a steel product for some special purposes,¹³⁾ such as ball bearings (15 μm), wire (20 μm), cans (20 μm) and tire cord (10 μm). Therefore, removing the small inclusions becomes an important goal, as most of the big inclusions (larger than 50 μm) can now be removed by highly developed traditional flow controllers.

Some investigations^14,15,16) showed that the small inclusions which are non-wetted with the liquid can be brought to the top surface by attaching to gas bubbles, or by being captured within a rising bubble’s wake. Besides, the upward flow caused by bubbles floating out of the bath, homogenizes liquid steel temperatures and composition. As such, gas bubbling is widely accepted^{17,18,19,20,21,22,23,24)} as an effective method for deeply cleaning the liquid steel in continuous casting operations. Zhang et al.²⁵⁾ reviewed the mechanisms of inclusion removal from molten steel by bubble flotation. Chang et al.²⁶⁾ performed a water modeling test in a seven-strand tundish, using submerged gas bubbling at various positions. They claimed that submerged gas bubbling from a suitable location on the bottom surface can lead to more uniform flow characteristics and temperatures at outlet ports of a tundish. Chattopadhyay et al.²⁷⁾ investigated the behavior of the “exposed eye” of slag layer in a gas shrouded tundish, using water modeling and a numerical simulation. They found that the area of the “slag eye” increases with rising gas flow rate. Wang et al.²⁸⁾ developed a mathematical model to predict inclusion removal by the flotation of bubbles of different sizes. The results showed that the optimum bubble size for the removal of small inclusions is predicted to be in the range of 0.5–2 mm in diameter in liquid steel. Generally speaking, bubbles generated from a porous plug are around 3 mm in diameter,²⁹⁾ in water, and are expected to be much larger (e.g. 10–50 mm in diameter) in liquid steel,³⁰⁾ owing to non-wetting effects and much higher surface tensions. In contrast with larger bubbles, micro-bubbles, defined as being smaller than one millimeter in diameter, have a higher total surface area per unit gas flow rate and a much longer residence time. Both factors should benefit inclusion removal.

In the present work, water model experiments were performed in a full scale water model tundish, in order to study inclusion behavior under various bubbling conditions. Inclusions in steel were modeled by using hollow spheres of borosilicate glass, close to the sub-50 micron size inclusions that need to be removed. Micro-bubbles were generated in a specially designed ladle shroud. The inclusions flowing out of the SENs (Submerged Entry Nozzle) to each strand were detected and recorded using the on-line Aqueous Particle Sensor equipment (APS III). The optimum gas injection scheme for the present equipment was obtained through analyzing the removal ratios of inclusions under different bubbling conditions.

2. Water Modeling

Water modeling was carried out using a full scale, delta-shaped, four-strand, acrylic model tundish, located in the water modeling laboratory of the MMPC (McGill Metals Processing Centre). The prototype of this model is a 12 t capacity tundish for a four-strand billet caster. The advantage of a full scale model has been previously addressed by Guthrie and Isac.³¹⁾ Thus, water is a suitable liquid to replace liquid steel at 1823 K, because the kinematic viscosities of the two are comparable, and the process is governed by Froude and Reynolds modelling criteria. As shown in Fig. 1, the set-up consists of the full-scale tundish, a 3 m high upper tank simulating the ladle, a ladle shroud and slide gate, a gas injection system, an inclusion supply system, an on-line Aqueous Particle Detection system and various flow meters. As such, we can expect velocity fields and turbulence to be exactly matched between the full scale model tundish, and that in the steel plant.

Fig. 1.

Full-scale water model of the delta shaped, four-strand tundish.

The configuration and key dimensions of the tundish are shown in Fig. 2. The inflow rate was maintained as 170 L/min with a slide gate opening of 39%, water flowing out from each strand at 42.5 L/min, in order to obtain a stable water level of 550 mm in the tundish, corresponding with plant practice. The ladle shroud was immersed in the tundish bath to a depth of 60 mm to stabilize the top surface.

Fig. 2.

Top view of the delta shaped tundish profile with key dimensions (mm).

2.1. Generation of Micro-bubbles

In a hydrostatic wetting system, the bubble size at low gas flow rates is solely determined by a balance between the buoyancy force and the surface tension force, which fixes the bubble to the inner wall of the nozzle. However, liquid steel is non-wetting to refractory SENs, whilst its surface tension is approximately twenty times greater than that of water. Hence, blowing gas into liquid steel through a bottom refractory porous plug causes the forming bubbles to spread and coalesce with each other over the surface of the plug, leading to much larger bubbles. As a consequence, the sizes of bubbles formed by bottom blowing into liquid steel will be much larger than those in a water model tundish, and therefore unrepresentative. In the present study, a different methodology was applied, injecting gas into the water flowing down through the ladle shroud. The crossflow of water (steel) can limit the sizes of bubbles forming, and can be expected to mitigate the impact of non-wetting behavior. As shown in Fig. 3, a specially designed ladle shroud was built for generating the micro-bubbles in the full scale water model. At the top of the ladle shroud, tiny orifices were drilled into the ladle shroud, in three layers located at 42 mm, 62 mm and 82 mm below the slide gate, respectively. Each gas injection layer contained four orifices, all at 90° to each-other. In previous modeling work at the MMPC,³²⁾ performed using a simplified inclined rectangular tank, it was determined that a smaller orifice is ideal for forming micro-bubbles in water. So, for this study, the diameter of each orifice was 0.3 mm, precisely drilled by laser. The input air was controlled by a precise needle valve, for different flow rates. The air flow rates to the ladle shroud were monitored, in-line, using a high precision thermos-gas flow rate meter, installed on the air line.

Fig. 3.

A view of the novel ladle shroud.

For bubble measurements, a plexiglass plate was set beneath the ladle shroud, with a 15 degree slant angle to the vertical. Once the micro-bubbles were generated, they would be entrained in the entry flow, and impact on the plate set beneath. A high speed camera was used to capture the images of the micro-bubbles hitting the plate. Intense background lighting was provided by two LED lights, for the high shutter speed of 1/2500 s. The micro-bubble images were post-processed using the open source software, “ImageJ”. The outputs provide major and minor axes of each bubble, as well as the average diameter of micro-bubbles. The bubble measurements and inclusion detection measurements were independent operations.

2.2. Inclusion Injection

Inclusions in steel were simulated by small, hollow, borosilicate glass, microspheres. All these microspheres were below 100 μm, so as to simulate the removal of non-metallic inclusions (especially those smaller than 50 μm) in industrial operations. The apparent density of the borosilicate particles was determined to be 700 kg/m³. In order to simulate actual industrial operations, the particles were stirred vigorously in water, forming a suspension of particles. This suspension was continuously injected into the top of the ladle shroud using a variable-speed peristaltic pump, at a fixed flow rate of 0.4 L/min. The concentration of the dispersed particles was 0.594 g/L, taking into account the water flow rate, the flow rate of the suspending liquid, and the density and size range of the microspheres. During the whole injection process, the suspension was stirred continuously, so as to maintain them submerged and well mixed. In actual fact, the average density of inclusions in the steelmaking process, which are mainly composed of alumina and silica particles, is ~3000 kg/m³. As the density ratios of the continuous phase to discrete phase are different in the two systems, (i.e. between the water modeling and industrial operations), similarity criteria were developed. Thus Sahai and Emi³³⁾ proposed the similarity criteria for different particles, based on the similarity of trajectories. As the present experiments were performed in a full scale water model, geometric similarity is not a consideration in this study. So, the similarity relation between the diameter of inclusions (d_i) and simulating particles (d_p) can be simplified to;   

$$\frac{d_{\mathrm{i}}}{d_{\mathrm{p}}}=\sqrt{({\gamma }_p-1)/({\gamma }_i-1)}$$(1)

where γ_p represents the density of a particle divided by the density of water, and γ_i is the density ratio of an inclusion and liquid steel. The dimensional ratio (0.72:1) of inclusions in liquid steel equivalent to our hollow glass microspheres can be obtained by substituting the densities of the corresponding materials of two systems according to Eq. (1).

2.3. Inclusion Detection

In this study, the simulated inclusions were detected on-line, using the APS III, which is an aqueous equivalent of the LiMCA (Liquid Metal Cleanliness Analyzer), invented by Guthrie and Doutre in 1985.³⁴⁾ This system has already been widely applied in the aluminum industry,^35,36,37,38) and is now also developed for steelmaking, and for water modelling. As shown in Fig. 4, the APS III consists of a tube with a tiny orifice in it, a data acquisition system, and a signal converter program. Both the APS III and the Coulter Counter³⁹⁾ are based on the Electric Sensing Zone (ESZ) Principle for ionically conducting fluids. In its operating mode, the APS III continuously withdraws some of the bulk water, and any inclusions contained within it, through the ESZ orifice, for particle counting. An electrically (ionically) conducting fluid (tap water) maintains a constant electric current between the external copper cathode and internal copper anode, located on either side of the orifice, or ESZ, located near the bottom of the electrically insulating, glass tube. If no inclusion flows through the orifice, the electrical resistance between two electrodes is determined only by the conductive liquid, and the ESZ. When a (non-conducting) inclusion passes through the ESZ, it occupies a fractional volume, leading to a resistance change. At constant current, this is displayed as a voltage jump in the data acquisition system.

Fig. 4.

Electric sensing zone principle.

Since the height of the resistive pulse is related to the size of the particle, its size can readily be analyzed. According to Maxwell’s analysis,⁴⁰⁾ the resistance of the ESZ containing a non-conductive particle is   

$$R_{\text{ESZ-paricle}}={\rho }_{\text{eff}}\int \frac{dx}{A(x)}$$(2)

where A(x) represents the cross-sectional area of the ESZ, and ρ_eff is the effective electrical resistivity of the orifice space filled with liquid metal and a particle.

If the concentration of particles in the liquid is low, the frequency of particles passing through the ESZ is also sufficiently low, such that the particles do not interfere with each other when passing through the ESZ. The effective electrical resistivity can be calculated as follows.   

$${\rho }_{\text{eff}}={\rho }_{\mathrm{e}}\left[\frac{2{\rho }_{\mathrm{e},\mathrm{\hspace{0.17em} }\mathrm{p}}+{\rho }_{\mathrm{e}}+f({\rho }_{\mathrm{e},\mathrm{\hspace{0.17em} }\mathrm{p}}-{\rho }_{\mathrm{e}})}{2{\rho }_{\mathrm{e},\mathrm{\hspace{0.17em} }\mathrm{p}}+{\rho }_{\mathrm{e}}-2f({\rho }_{\mathrm{e},\mathrm{\hspace{0.17em} }\mathrm{p}}-{\rho }_{\mathrm{e}})}\right]$$(3)

where, ρ_e and ρ_e,p represent the electrical resistivity of the liquid and particles respectively. f is the volume fraction of the particle passing through the orifice.   

$$f=\frac{2d_{\mathrm{p}}{}^3}{3D^{2}L}$$(4)

where L is the length of the ESZ, and d_p and D, are the diameters of the particles and orifice. So the resistance caused by the particle can be represented as the difference between the resistance of the ESZ, with and without the particle.   

$$\mathrm{\Delta }R=\frac{4L}{\pi D^2}({\rho }_{\text{eff}}-{\rho }_{\mathrm{e}})$$(5)

The borosilicate glass tube, containing the ESZ, is an excellent insulator with an electrical resistivity in the range of 10¹¹−10¹³ Ω m at room temperature. By contrast, the electrical resistivity of water is several orders of magnitude smaller, (1.8*10⁵Ω m). Hence, an approximate expression of the effective electrical resistivity is   

$${\rho }_{\text{eff}}=\frac{1}{2}{\rho }_{\mathrm{e}}\left(\frac{2+f}{1-f}\right)$$(6)

Then, converting the above equation into the power series expansion format, as follows;   

$${\rho }_{\text{eff}}={\rho }_{\mathrm{e}}\left(1+\frac{3}{2}f+\frac{3}{2}{f}^2+\cdot \cdot \cdot \right)$$(7)

The particles are quite small compared with the orifice diameter, so that only the first two terms are taken into consideration, and the impact of higher order terms can be ignored. So, when a small particle passes through the orifice, the instantaneous resistance change can be represented as follows.   

$$\mathrm{\Delta }R=\frac{4{\rho }_{\mathrm{e}}{d}_{\mathrm{p}}{}^3}{\pi D^4}$$(8)

and the corresponding transient value of voltage pulse is:   

$$\mathrm{\Delta }V=\frac{4{\rho }_{e}I{d}_{\mathrm{p}}{}^3}{\pi D^4}$$(9)

where I is the imposed current. As such, the size of the particle can be deduced from the value of the peak voltage pulse.

The driving force on particles passing through the orifice of the APS III device was provided by employing a syphon system. Thus, instantaneous fluctuations in the suction rate can seriously affect the stability of the measuring signal, causing a large baseline noise band on the screen, when using an electrical pump. The syphon provides negative pressure through the height difference of liquid levels, which made the suction rate precisely controllable, while providing a noise free system. The syphon tube was controlled via a small valve, keeping the suction rate around 7 ml/min and the corresponding flow velocity in orifice at 1.65 m/s.

It is worth noting that the relationship between the voltage pulse and the particle size described in Eq. (9) is just a theoretical deduction under completely ideal conditions. Therefore, the APS device needed to be calibrated before being applied in the experiments. Calibration was performed separately using four sizes of standard particles, namely 20, 40, 90 and 140 μm. These four kinds of standard particles can be described in a coordinate system using their sizes as x-coordinates and the measured voltages as the y-coordinates, forming four points, in addition to the zero at zero value. Figure 5 shows the fitted curve for these points, plus the origin, using the data analysis software of OriginLab. The final formula of the relation between the voltage and particles size could be expressed as a polynomial   

$$y=8.23\times {10}^{-7}{x}^3-1.78\times {10}^{-5}{x}^2+7.56\times {10}^{-3}x$$(10)

For easy recording, the whole measuring range was divided into eleven voltage intervals. When a particle records a voltage signal located in a particular voltage interval, its size is regarded as being the mean size calculated from the two marginal voltages of the corresponding step. Table 1 lists the voltage steps, the corresponding particle size range, the mean sizes, plus the equivalent size of a corresponding inclusion in liquid steel.

Fig. 5.

Relationship between voltage and inclusion size.

Table 1. Inclusion detection algorithm.

Bin No.	Voltage range (mV)	Particle size range in water (μm)	Mean size (μm)	Equivalent inclusion size in liquid steel (μm)
1	0.15 to 0.20	19.92 to 26.09	23.01	16.56
2	0.20 to 0.30	26.09 to 37.30	31.70	22.82
3	0.30 to 0.40	37.30 to 46.85	42.08	30.29
4	0.40 to 0.50	46.85 to 55.07	50.96	36.69
5	0.50 to 0.60	55.07 to 62.23	58.65	42.43
6	0.60 to 0.80	62.23 to 74.20	68.22	49.11
7	0.80 to 1.00	74.20 to 84.11	79.16	56.99
8	1.00 to 1.50	84.11 to 103.34	93.73	67.48
9	1.50 to 2.00	103.34 to 118.06	110.70	79.70
10	2.00 to 3.00	118.06 to 140.60	129.33	93.12
11	3.00 to 5.00	140.60 to 172.47	156.54	112.71

2.4. Experimental Procedures

After attaining steady state flows within the full scale, water model, tundish, the gas supply system was activated so as to generate micro-bubbles. The concentrated suspension, composed of particles and water, was kept in a big graduated cylinder, in order to provide the same number density of particles for each test. A variable-speed peristaltic pump was used to suck the suspension from the cylinder and inject it into the ladle shroud. Before the start of each measurement, the particle solution was pre-injected for ten minutes, in order to reach steady state conditions for the flow of particles in the tundish. This corresponds to a Residence Time of θ =1.4, given a Nominal Residence Time (i.e. θ=1), of 7 minutes. DASYLab, a signal acquisition system, began to record the voltage pulses caused by particles passing through the ESZ. At the same time, the rate of liquid suction was recorded during the recording of passing particles. Figure 6 shows the working interface of the DASYLab. To simplify the experiments, particles were detected at only two SENs, 3 and 4, on the right side of the centerline, as symmetrical flow of water on the two sides of the entering pouring stream could be assumed.

Fig. 6.

The working interface of the data acquisition system.

3. Results and Discussion

Three variables: a). gas injection position, b). port number and c). gas flow rate, were considered in the experiments to optimize bubble conditions. The outflow particle number density (n_p) describes the particle content in unit volume of fluid exiting the tundish from each SEN.   

$$n_{\mathrm{p}}=\frac{N_{\mathrm{p}}}{V_s}$$(11)

Here N_p represents the total number of particles detected during the measuring process, and V_s represents the fluid volume sucked out by the sensor during the same time period.

3.1. Position of Gas Injection

The gas injection position affects bubble sizes owing to the non-uniform distributions of flow velocities and local dissipation rates of turbulence (ε) within the ladle shroud. Three different injecting positions were used in this study. The first, second, and third gas injection layers were located 42 mm, 62 mm, and 83 mm, below the bottom of the slide gate, respectively. In each case, gas was injected through one port at a fixed gas flow rate of 0.2 L/min, in order to leave the injection position as the only variable.

As shown in Fig. 7, the mean diameter of the bubble is 814 μm when the gas was injected from the port located at the first gas injection layer. The average bubble size grows to 895 and then to 915 μm for the lower injection ports located on the second and third injection layer, respectively. Since the slide gate was not-fully opened in this study, the turbulent kinetic energy and the turbulent dissipation rate (ε) also decay downstream of the ladle shroud. Under highly turbulent flow conditions, the bubbles formed can be distorted and broken by turbulent stresses, if the turbulence is strong enough to reach a Critical Weber Number.⁴¹⁾ The splitting of bubbles by turbulent flow was investigated by Hinze,⁴²⁾ and a Critical Weber Number of 1.2 was attained from their experiments.   

$$W{e}_c=\frac{{\rho }_{\mathrm{w}}{C}_1{\epsilon }^{2/3}{d}_{\mathrm{b}}{}^{5/3}}{\sigma }=1.2$$(12)

Here ρ_w is the density of water, σ is the interfacial surface tension, and C₁ is the empirical constant which is equal to 2. So the relationship between bubble diameter and the turbulent dissipation rate could be calculated and this is shown in Fig. 8.

Fig. 7.

The bubble sizes generated from different gas injection positions: (a) the first layer, (b) the second layer, (c) the third layer, for a fixed gas flow rate of 0.2 L/min, with single injection port.

Fig. 8.

Critical bubble size under for different dissipation rates of turbulence, ε.

A simple numerical simulation was performed to show the distribution of dissipation rates of turbulent kinetic energies in the ladle shroud (shown in Fig. 9). For simplifying the computation, only 15 cm length of ladle shroud below the slide gate, is shown in this study. It is obvious that the closer the gas injection position is to the slide gate, the higher is the turbulent dissipation rate. This can potentially break up the initial forming bubbles, into smaller ones. The predicted turbulent dissipation rates were around 90, 40 and 30 m²/s³, near to the ports at the first, second, and third gas injection layers, respectively. These corresponded to predicted bubble sizes of 403, 557, 625 μm (based on Eq. (12)), which were smaller to measured bubble sizes. In this study, the forming bubbles passed very rapidly through the turbulent flow region close to the slide gate nozzle, where the kinetic energy of turbulent dissipation was greatest. Therefore, bubble break-up due to turbulent fluctuations was limited by their short transit times.

Fig. 9.

The dissipation rate of turbulence in non-fully opened ladle shroud. (Online version in color.)

Figure 10 is the outflow particle number density for the entire tundish, for different gas injection positions, as measured by APS III. The graph shows an obvious declining trend with decreasing distance between the injection port and the slide gate, indicating enhanced removal of particles. The length of the error bars represent the standard deviations based on four identical experimental tests. The sizes of bubbles were reduced, when the injection port was located closer to the slide gate. According to the fundamentals of inclusion removal by bubble flotation, smaller bubbles have a better performance for inclusion removal, as compared with bigger bubbles. For a given net gas flow rate, smaller bubbles can provide a larger total surface area for inclusions to become attached. Furthermore, smaller bubbles have a longer residence time within the tundish, which increases the probability of collision between inclusions and bubbles.

Fig. 10.

The averaged outflow particle number density from the tundish using different gas injection positions.

Table 2 presents all the detailed results, including the bubble sizes, and the particle number densities, obtained from the inner strand, the outer strand, and the average for the whole tundish. When the gas injection position was raised from the third layer to the first layer, the outflow particle number density of the entire tundish only dropped by 7.53%, reducing from 11.755/ml to 10.87/ml. This means that the injection position is not a key factor for improving the inclusion removal. That is mainly because the three injection layers are quite close to each other in this experiment. However, a certain correlation still exists between gas injection position and n_p. For minimizing bubble sizes, all the activated ports were located in the first injection layer for the following tests. It is notable that the n_p obtained from the inner SEN exit port was around two to three times higher than that for the outer nozzle. This is mainly because the outer SEN nozzles, 1 and 4, are further from ladle shroud than are the inner strands, 2 and 3, so that the longer residence time before liquid reaches the outer strands allows particles (inclusions) to enjoy a longer time and a greater probability to float up to the top surface, before flowing out of the exit outer SENs.

Table 2. The outflow particle number densities (n_p) and average bubble size under different gas injection positions.

Gas flow rate (L/min)	Port number	Port layer	Mean bubble diameter (μm)	np of Inner Strand (/ml)	np of Outer Strand (/ml)	Average np (/ml)
0.2	1	1	814	15.30	6.44	10.87
0.2	1	2	895	16.07	6.92	11.495
0.2	1	3	915	16.49	7.02	11.755

3.2. Number of Ports

Multi-port gas injection in the ladle shroud was used to generate as small bubbles as possible, while providing for a sufficient number of bubbles to be used within the tundish for particle removal. Figure 11 shows the bubbles generated by single-port, double-port and four-port injection, for a fixed gas flow rate of 0.2 L/min. As mentioned above, all the operational ports were located in the first layer.

Fig. 11.

The bubbles formed by (a) single-port gas injection, (b) double-port gas injection, (c) four-port gas injection, for a gas flow rate of 0.2 L/min, all ports located at the first layer.

According to the post-processing results of these bubble pictures, the average bubble size was 814 μm in diameter with the single-port gas injection. When gas was introduced by double-port injection, the mean size of bubbles decreased to 727 μm, a drop of 10.7%. Then, it was further reduced to 633 μm, with four-port gas injection. Therefore, the mean bubble sizes were reduced with an increase in gas injection ports, for the same overall gas flowrate.

As shown in Fig. 12, the results of APS detection illustrates that the averaged outflow particle number density follows the same tendency with bubble size under different injection port numbers. The n_p of the entire tundish was 10.87/ml when bubbles were generated through single port gas injection. With double-injection, it declined by 17.2% to 9.00/ml and further reduced to 7.85/ml, another 10.6% drop, when the active port number increased to four. Under a constant total gas flow rate, increasing the number of gas injection ports is equivalent to reducing the gas flow rate to each port, which is beneficial in generating smaller bubbles.

Fig. 12.

The averaged outflow particle number density for different gas injection ports numbers.

More details are listed in Table 3. With increasing number of ports, the particle number densities of inner strand reduced from the 15.3/ml to 10.24/ml, a drop of 33.1%, while those detected at outer strand dropped by only 15.2%, being decreased from 6.44/ml to 5.46/ml. The removal of particles from the inner strand was significantly improved, versus that for the outer strand. Admittedly, micro-bubbles have a much wider spatial distribution within a tundish than that for larger bubbles typically found (3–6 mm). Nevertheless, they were still concentrated relatively close to the ladle shroud entry region. So the improvement in the removal of particles was more obvious for the inner strand.

Table 3. The outflow particle number densities (n_p) and average bubble size under various gas injection port numbers.

Gas flow rate (L/min)	Port number	Port layer	Mean bubble diameter (μm)	np of Inner Strand (/ml)	np of Outer Strand (/ml)	Average np (/ml)
0.2	1	1	814	15.30	6.44	10.87
0.2	2	1	727	12.05	5.95	9.00
0.2	4	1	633	10.24	5.46	7.85

It is certain that increasing the number of injection ports is beneficial for inclusion removal. However, this tendency is not linear. This improvement is gradually weakened with an increasing number of ports. Due to the limited space within the ladle shroud, the coalescence between bubbles becomes more serious as the numbers of bubbles are increased. Owing to the superiority of multi-port gas injection, four-injection, the maximum port number of one layer in the present model, was applied in the following tests for different gas flow rates.

3.3. Gas Flow Rate

In this study, four specific gas flow rates, between 0.1 to 0.8 L/min, were injected through four-injection ports at the first port layer, in order to analyze the influence of gas flow rate on inclusion removal.

As shown in Fig. 13, the bubble size is reduced with decreasing gas flow rate. When increasing the gas flow rate, the gas flow velocity from the injection ports is increased. Since the turbulence formed by the inflow velocity and the degree of opening of the slide gate nozzle was fixed in this study, it seems insufficient for breaking the faster gas stream into tiny bubbles. Therefore, a low gas flow rate is recommended for the generation of small bubbles. This is also validated by the results of post-processing.

Fig. 13.

The bubbles introduced from the first injection layer with four-port gas injection, for gas flow rates of (a) 0.1 L/min, (b) 0.2 L/min, (c) 0.4 L/min and (d) 0.8 L/min.

As listed in Table 4, the mean size of bubbles was 543 μm in diameter for a gas flow rate of 0.1 L/min. It increased by 50.1%, up to 937 μm, when gas flow rate rose from 0.1 L/min to 0.8 L/min. However, with the increasing gas flow rate, the effect of particle removal presents a different trend from bubble size. Figure 14 shows the relationship between average particle number densities of the exit flows, versus the gas flow rate into the ladle shroud. The n_p was 8.215/ml under the gas flowrate of 0.1 L/min. With a rising of gas flow rate, this first reduced to 7.85/ml for a gas flow rate of 0.2 L/min, but then increased to 8.63/ml, and later rose to 11.515/ml for a gas flow rate of 0.8 L/min. It is clear that the bubbles grow with an increasing gas flow rate, which goes against efficient removal of inclusions. Nonetheless, a sufficient quantity of bubbles is also a key requirement for good inclusion removal, and this needs to be supported by a relatively higher gas flow rate. Therefore, the gas flow rate needs to be kept at an appropriate level to balance the bubble size and the quantity of bubbles, rather than as low as possible. As a result, the optimum gas flow rate for particle removal was 0.2 L/min, a medium level among all the gas flow rates. Consequently, the optimum bubbling condition for this model was gas injection through four ports located on the first injection layer for a total gas flowrate of 0.2 L/min.

Table 4. The particle number densities (n_p) of outflow under different gas flow rates.

Gas flow rate (L/min)	Port number	Port layer	Mean bubble diameter (μm)	np of Inner Strand (/ml)	np of Outer Strand (/ml)	Average np (/ml)
0.1	4	1	543	11.55	4.88	8.215
0.2	4	1	633	10.24	5.46	7.85
0.4	4	1	845	11.41	5.85	8.63
0.8	4	1	937	14.47	8.56	11.515

Fig. 14.

The averaged outflow particle number density for different gas flow rates.

Furthermore, for comparison, particle detection under no-gas conditions was also performed by closing all the gas injection ports. The particle number densities of inner and outer nozzles in bare tundish were 16.54/ml and 9.02/ml, respectively, and the value for the entire tundish was 12.78/ml. With gas bubbling, the worst result was obtained from single port gas injection at the third layer with a gas flow rate of 0.2 L/min, giving a particle number density in the outflow of 11.755/ml. This was only 8.02% less than that for a bare tundish. However, with the optimum bubble condition, a 38.6% improvement in particle removal was attained, compared with the bare tundish.

3.4. Particle Detection of the Inlet Stream

In contrast with outflow particle number density, the residual ratio of inclusions (RRI) is a more practical and intuitive criterion for evaluating the efficiency of each bubble condition, which can be represented as follows.   

$$RRI=\frac{\mathrm{\sum }_{i=1}^4(n_{\text{out},\text{i}}{Q}_{\text{out},\text{i}})}{Q_{\mathrm{w}}{n}_{\text{in}}}$$(13)

where n_in,i and n_out,i represent the particle number density of the inlet and each SEN, respectively. Q_w and Q_out,i are the water flow rate of entry flow and each nozzle. For RRI computations, the inflow particle number is necessary. Unfortunately, it is difficult to detect the particles in the inlet stream. The strong turbulent flow in the inlet duct disturbs the suction rate of the syphon tube, which causes serious signal interferences on the data acquisition system of APSIII. But, as the injecting rate of the particle solution is fixed, the particle number density of inflow can be calculated through the following equation.   

$$n_{\text{in}}=\frac{Q_{\mathrm{p}}{N}_{\mathrm{p}}}{Q_{\mathrm{w}}{V}_{\mathrm{p}}}\times {10}^{-6}$$(14)

where Q_p represents the flow rate of the particle solution. The total volume of particle solution is expressed as V_p. As multi-size particles were used in this study, the total number of particles in the solution, N_p, can be obtained from   

$$N_{\mathrm{p}}=\frac{m}{{\rho }_{\mathrm{p}}\sum p_{\mathrm{i}}\frac{1}{6}\pi d_{\mathrm{p},\mathrm{a}\mathrm{v}}{}^3}$$(15)

where m is the total mass of the particles; d_p,av represents the average diameter of particles in a given size range, and p_i represents the corresponding number fraction of these particles.

The size distribution of the particles was measured in a well-mixed particle solution using APS III, and is shown in Fig. 15. As seen, the highest proportion exists in the size range from 26.9 to 37.3 μm, accounting for 30%. More than 70% of the particles are smaller than 46.9 μm in diameter. Further, there are no particles larger than 103 μm, confirming the product description of the hollow glass microspheres. According to Eqs. (14) and (15), the particle number density of the inflow is 38.4/ml. Thus, the RRI under different bubble conditions for a bare tundish are listed in Table 5.

Fig. 15.

Size distribution of the particles.

Table 5. Residual ratio of inclusions for different gas injection schemes.

Case	Gas flow rate	Port number	Layer	RRI
1	0.1	4	1	21.39%
2	0.2	4	1	20.44%
3	0.4	4	1	22.47%
4	0.8	4	1	29.99%
5	0.2	2	1	23.44%
6	0.2	1	1	28.31%
7	0.2	1	2	29.93%
8	0.2	1	3	30.61%
Bare tundish	0	0	0	33.28%

For a well-designed tundish, the RRI of the bare tundish was already as low as 33.28%. When injecting the gas into the ladle shroud, the cleanliness of liquid was enhanced to varying degrees, depending on different gas injection schemes. Under the optimum gas bubbling condition, the RRI was improved dramatically, being reduced to 20.4% from 33.3%, equivalent to a 39% improvement. Figure 16 shows the residual particle number densities for each particle size interval, being detected in a bare tundish and the corresponding result when using micro-bubble swarms. It reveals that the efficiency of micro-bubbles was mainly focused on the removal of inclusions smaller than 50 μm in diameter. Figure 17 compares the inlet and outlet particle number densities in a bare tundish with, and without, micro-bubble swarms. It can be observed that the well-designed tundish was effective in removing the larger inclusions (greater than 50 μm), while the presence of micro-bubbles succeeded in removing many of the smaller inclusions (below 50 μm).

Fig. 16.

The average outlet particle number densities for different particle size intervals, for a bare tundish with and without micro-bubble swarms.

Fig. 17.

The particle number densities for a) inlet and for outlets; b) a bare tundish and c) a bare tundish with micro-bubbles.

However, it is noteworthy that there are still one fifth of micro-particles left in the outflow, even under the optimum bubbling condition. Since the present work only focuses on enhancing the inclusion removal by micro-bubbles, all the experiments were performed in the tundish without any flow modifiers. Therefore, the appearance of short circuit flows is inevitable. Compared with the adhesion force between bubbles and inclusions, the short circuit flows to exit ports 2 and 3 is strong and could cause inclusions to become detached from the bubbles. As such, some particles affected by short circuit flows may possibly not be removed by micro-bubbles.

4. Summary

The present work focused on the generation and effect of micro-bubbles on inclusion removal in a bare tundish. Water modeling was performed in a full-scale, delta-shaped, four-strand water model tundish. In order to enhance the inclusion removal, micro-bubbles were generated by a specially designed ladle shroud with 12 laser-drilled orifices (0.3 mm in diameter). Hollow glass particles, smaller than 100 μm in diameter, were used to simulate the non-metallic inclusions in liquid steel, and detected by APS III, a novel particle sensor. The main conclusions were drawn as follows;

1. Inclusion removal was enhanced by gas bubbling into the ladle shroud. The effect of gas bubbling on inclusion removal is very dependent on the gas injection scheme. The lowest residual ratio of inclusions (20.44%) was achieved when gas was injected through four ports within the first injection layer, at a gas flow rate of 0.2 L/min.

2. There is a clear correlation between gas injection position and its effect on inclusion removal, the closer the injection port to the slide gate was, the better it performed in removing inclusions.

3. Multi-port gas injection is an effective method to promote the inclusion removal, as it dramatically reduces the gas flow rate to each port, while keeping the total gas flow rate constant. Certainly, increasing the port number is beneficial for inclusion removal.

4. Under a given method of gas injection, the optimum gas flow rate is an intermediate value, determined by balancing smaller bubble sizes against the total number of bubbles being created.

5. Due to the lack of tundish flow controllers to enhance “steel quality”, one-fifth of micro-particles exit as a result of short circuiting flows, and could not be removed by micro-bubbles.

A(x): cross-sectional area of ESZ (m²)

d_i, d_p, d_b: diameter of the inclusion, particle and bubble (m)

d_pav: average diameters of particles in a certain size range (m)

D: orifice diameter of the APS tube (m)

f: volume fraction of the particle in ESZ (-)

I: electric current (A)

L: length of the electric sensing zone (m)

m: total mass of particles (kg)

n_p: particle number density (number/ml)

N_p: number of particles (-)

p_i: number fraction of particles in a certain size range (-)

Q_w, Q_out,i, Q_p: flow rates of entry flow, outlet i and particle suspension (m³/s)

ΔR: resistance change of the SEZ cause by particle (Ω)

R_ESZ-particle: resistance of the ESZ with particles (Ω)

V_s: sucking volume of liquid (ml)

V_p: total volume of particle suspension (m³)

ΔV: voltage jump caused by particle (V)

γ_p: density ratio of a particle and water (-)

γ_i: density ratio of a inclusion and liquid steel (-)

σ: surface tension coefficient of water (N/m)

ε: turbulent dissipation rate (m²/s³)

θ: residence time/nominal residence time (-)

ρ_e, ρ_e,p: effective electrical resistivity of liquid and particles (Ωm)

ρ_eff: effective electrical resistivity (Ωm)

ρ_p, ρ_w: densities of particle and water (kg/m³)

ESZ: Electric Sensing Zone

SEN: Submerged Entry Nozzle

RRI: Residual Ration of Inclusions

APS: Aqueous Particle Sensor

References

• 1)   R. D.  Morales,  S.  Lopez-Ramirez,  J.  Palafox-Ramos and  D.  Zacharias: ISIJ Int., 39 (1999), 455.
• 2)   R.  Schwarze,  D.  Haubold and  C.  Kratzsch: Ironmaking Steelmaking, 42 (2015), 148.
• 3)   T.  Merder and  M.  Warzecha: Metall. Mater. Trans. B, 43B (2012), 856.
• 4)   A.  Tripathi and  S. K.  Ajmani: ISIJ Int., 45 (2005), 1616.
• 5)   S. G.  Zheng and  M. Y.  Zhu: Ironmaking Steelmaking, 33 (2006), 398.
• 6)   S.  Sing and  S. C.  Koria: Ironmaking Steelmaking, 23 (1996), 255.
• 7)   H. B.  Kim,  M.  Isac and  R. I. L.  Guthrie: Steel Res. Int., 76 (2005), 53.
• 8)   K.  Raghavendra,  S.  Sarkar,  S. K.  Ajmani,  M. B.  Denys and  M. K.  Singh: Appl. Math. Model., 37 (2013), 6284.
• 9)   A.  Cwudzinski: Steel Res. Int., 85 (2014), 902.
• 10)   L. C.  Zhong,  R. C.  Hao,  J. Z.  Li and  Y. X.  Zhu: J. Iron Steel Res. Int., 21 (2014), 10.
• 11)   S. A.  Morsi and  A. J.  Alexander: J. Fluid. Mech., 55 (1972), 193.
• 12)   H. V.  Atkinson and  G.  Shi: Prog. Mater. Sci., 48 (2003), 457.
• 13)   L. F.  Zhang and  B. G.  Thomas: ISIJ Int., 43 (2003), 271.
• 14)   L. T.  Wang,  Q. Y.  Zhang,  C. H.  Deng and  Z. B.  Li: ISIJ Int., 45 (2005), 1138.
• 15)   J. P.  Rogler,  L. J.  Heaslip and  M.  Mehrvar: Can. Metall. Q., 44 (2005), 357.
• 16)   D.  Reay and  G. A.  Ratcliff: Can. J. Chem. Eng., 51 (1973), 178.
• 17)   D. F.  Chen,  X.  Xie,  M. J.  Long,  M.  Zhang,  L. L.  Zhang and  Q.  Liao: Metall. Mater. Trans. B, 45B (2014), 392.
• 18)   K.  Chattopadhyay,  M.  Isac and  R. I. L.  Guthrie: ISIJ Int., 51 (2011), 759.
• 19)   A.  Ramos-Banderas,  R. D.  Morales,  L.  Garcia-Demedices and  M.  Diaz-cruz: ISIJ Int., 43 (2003), 653.
• 20)   A.  Vargas-Zamora,  R. D.  Morales,  M.  Diaz-Cruz,  J.  Palafox-Ramos and  J. D. J.  Barreto-Sandoval: Metall. Mater. Trans. B, 35B (2004), 247.
• 21)   R.  Liu and  B. G.  Thomas: Metall. Mater. Trans. B, 46B (2015), 388.
• 22)   J. P.  Rogler,  L. J.  Heaslip and  M.  Mehrvar: Can. Metall. Q., 43 (2003), 407.
• 23)   N.  Kubo,  T.  Ishii,  J.  Kubota and  T.  Ikagawa: ISIJ Int., 44 (2004), 556.
• 24)   T.  Zhang,  Z. G.  Luo,  L. C.  Lin,  H.  Zhou and  Z. S.  Zou: Powder Technol., 273 (2015), 154.
• 25)   L. F.  Zhang and  S.  Taniguchi: Int. Mater. Rev., 45 (2000), 59.
• 26)   S.  Chang,  L. C.  Zhong and  Z. S.  Zou: ISIJ Int., 55 (2015), 837.
• 27)   K.  Chattopadhyay,  M.  Isac and  R. I. L.  Guthrie: ISIJ Int., 51 (2011), 573.
• 28)   L. H.  Wang,  H. G.  Lee and  P.  Hayes: ISIJ Int., 36 (1996), 7.
• 29)   K.  Chattopadhyay,  M.  Hasan. M. Isac and R. I. L. Guthrie: Metall. Mater. Trans. B, 41B (2010), 225.
• 30)   G. A.  Irons and  R. I. L.  Guthrie: Metall. Trans. B, 9B (1978), 101.
• 31)   R. I. L.  Guthrie and  M.  Isac: ISSTech Conf. 2003, AIST, Warrendale, PA, (2003), 1201.
• 32)   X. Y.  Ren: Master Thesis, McGill University, (2015), 1, 2.
• 33)   Y.  Sahai and  T.  Emi: ISIJ Int., 36 (1996), 1166.
• 34)   D.  Doutre and  R. I. L.  Guthrie: Method and Apparatus for the Detection and Measurement of Particulates in Molten Metal, U. S. Patent, 4555662, (1985).
• 35)   M.  Li and  R. I. L.  Guthrie: Metall. Mater. Trans. B, 31B (2000), 767.
• 36)   R. I. L.  Guthrie and  M.  Li: Metall. Mater. Trans. B, 32B (2001), 1081.
• 37)   R. I. L.  Guthrie and  H. C.  Lee: Proc. 75th Steelmaking Conf., ISS-AIME, Warrendale, PA, (1992), 799.
• 38)   X. D.  Wang,  M.  Isac and  R. I. L.  Guthrie: ISIJ Int., 49 (2009), 975.
• 39)   W. H.  Coulter: U. S. Patent, 2656508, (1953), Means for Counting Particles Suspended in a Fluid.
• 40)   J. C.  Maxwell: A Treatise on Electricity and Magnetism, 3rd ed., Vol. 1, Clarendon, Oxford, (1954), 429.
• 41)   G. M.  Evans,  G. J.  Jameson and  B. W.  Atkinson: Chem. Eng. Sci., 47 (1992), 3265.
• 42)   J. O.  Hinze: AIChE J., 1 (1955), 289.