Fluid flow in converter bath, affecting the viability, effectiveness, and efficiency of steelmaking, plays a critical role in the productivity and quality level that can be achieved in the process. Due to a large quantity and diversity of the studies on the characteristics of the multiphase flow, it seems very necessary to make a systematic literature review on state-of-the-art developments in the steelmaking process. This paper presents the recent findings of the characteristics of the multiphase flow in the converter by means of physical and mathematical modeling and the resulting implications for simulating the process. Some representatives include supersonic oxygen jet, stirring and mixing, splashing and droplet generation, and energy transfer. The work summarized in this paper can give an in-depth understanding of the fluid flow in converter and provide references for future modeling of the converter steelmaking process. Future contributions to a fundamentally generalized modeling of the converter steelmaking are still needed. More profoundly, the modeling work can facilitate the real-time data-driven precise BOF process control and be an important component to the realization of intelligent manufacturing in steelmaking process.

The reduction of metal oxides by carbon monoxide is a very significant reaction in metallurgical and chemical industries. In current study, a kinetic model describing the reduction reaction of metal oxides by CO has been developed. The model is in the analytic function of parameters such as temperature, radius of the particle, CO content, density of metal oxides and time, which is, therefore, convenient for usage and theoretical analysis. The developed model can be applied to describe the reduction kinetics of metal oxides under both isothermal and non-isothermal conditions. Particularly, it can be used to describe the kinetics of both single reduction reaction (single reaction interface) and dual reduction reactions (double reaction interfaces, such as, the reduction of Fe₂O₃ to FeO with Fe₃O₄ being the intermediate product) as well. Application of the developed model to practical systems validated that its predicted values agreed very well with the experimental data in literatures.

As a part of ongoing efforts to develop a novel clean smelting process of Hongge vanadium titanomagnetite (HVTM), the effects of preheating temperature and time of HVTM pellet (HVTMP) on its gas-based direct reduction behavior with simulated shaft furnace gases were investigated in this paper. The results showed that the compressive strengths of preheated and roasted HVTMP could be improved by increasing the preheating temperature and time to some extent. The induration mechanism of HVTMP with different preheating conditions was further deeply explained. Increasing the preheating temperature and time of HVTMP accelerated its subsequent reduction rate, however, from 900 to 1000°C, the reduction rate decreased. The preheating temperature and time of HVTMP had little effect on the final phase composition of reduced HVTMP. A schematic illustration was proposed to describe the effects of preheating temperature and time of HVTMP on its subsequent reduction swelling. In the practical industrial production of HVTMP, the preheating characteristics should be optimized to meet the requirements of compressive strength and reduction swelling of HVTMP smelting in the gas-based shaft furnace. This study aims to provide both theoretical and technical basis for the effective utilization of HVTM either in blast furnace or shaft furnace.

This paper investigated the influence of Cr₂O₃ addition on the gas-based direct reduction behavior of Hongge vanadium titanomagnetite pellet (HVTMP) with simulated shaft furnace gases. The reduction curves obtained indicated that the Cr₂O₃ addition revealed different effects on the reduction of HVTMP and the reduction process could be divided into three stages. X-ray diffraction showed that FeCr₂O₄ was formed by adding Cr₂O₃, which was hard to be reduced and had an adverse effect on the metallization degree of final reduced HVTMP. The morphology analysis revealed that the connected metallic iron particles were destroyed and their growth and connection were prevented during the reduction when Cr₂O₃ was added. With the increase of Cr₂O₃ addition, the growth of metallic iron whiskers was changed form close cluster to disperse location, and the number of metallic iron whiskers increased, which determined the extent of reduction swelling. In addition, the reduction swelling had relationship with the compressive strength of reduced HVTMP with different Cr₂O₃ additions. This study will not only establish a relationship between the Cr₂O₃ addition of HVTMP and its gas-based direct reduction behavior, but also contribute greatly to the effective utilization of HVTM in shaft furnace.

The reserves of niobium and rare earth element in the mineral resources in Bayan Obo ranked the forefront of the world, the thermodynamic information such as liquidus and phase equilibrium relations in the related slag system phase diagram were important to the comprehensive utilization of niobium and rare earth resources. In the current work, the pseudo-melting temperatures were determined by the single-hot thermocouple technique (SHTT) for the specified content of 10 to 50 pct Nb₂O₅ in the CaO-SiO₂-Nb₂O₅-10%La₂O₃ phase diagram system. The 1573 K, 1623 K, and 1673 K liquidus were first calculated based on the pseudo-melting temperatures according to thermodynamic equations in the specific primary crystal area. The phase equilibrium relations at 1573 K and 1473 K were determined experimentally using the high-temperature equilibrium followed by scanning electron microscope, X-ray diffraction, and energy dispersive X-ray spectroscope analysis. The liquid phase, SiO₂ phase, CaO·SiO₂ phase, and 2CaO·Nb₂O₅ phase were found in the experiment. Therefore, the phase diagram was constructed for the specified region of the CaO-SiO₂-Nb₂O₅-10%La₂O₃ system. The experimental results have practical significance for further research on the related slag system, and for the comprehensive utilization of niobium and rare earth resources.

Impeller stirring is an important industrial technique which is used in hot metal pretreatment processes in steelmaking. The liquid/liquid flow characteristics of impeller stirring were clarified by a two-dimensional PIV experiment and CFD analysis and a liquid/liquid system consisting of liquid paraffin (oil) and ion-exchanged water (water). The target mixing patterns were Types I, II and III. Here, Type I is defined as the region where the two liquid phases exist separately, Type II is the region where dispersed oil exists in a scattered form in a continuous water phase, and Type III is the region where dispersed air and oil existed in the continuous liquid phase of water. The calculated oil/water and air/water interfaces were in good agreement with the experimental results. The vertical upward flow near the wall resulted in one or two circulation flows, one near the oil/water interface and the other below it, whereas the downward flow formed a single circulation flow near the bottom. The cross-sectional mean water velocity and turbulence energy displayed their peak values at the blade side regardless of the mixing patterns. In addition to the large value at the blade side, turbulence energy displayed a large value near the air/water interface in the type III mixing pattern. In Types I and II, the cross-sectional mean water velocity and turbulence energy near the oil/water interface increased at higher rotation speeds. The turbulence energy at the gap between the water and the side of the paddle blade proceeded toward the vessel wall.

The carbothermic reduction experiments were carried out for vanadium-titanium magnetite in alkaline molten in argon atmosphere at high temperatures. The effects of reduction temperature, carbon content and addition agent on the formation of pig iron containing vanadium were studied by X-ray diffraction (XRD) and scanning electron microscope (SEM). The XRD patterns of reduced slags results showed that Fe phase disappeared and the main phase of the reduced sample were Na₁₆Ti₁₀O₂₈, CaTiO₃, and Na_1.66AlSiO_4.33 when the reduction temperature was more than 1473 K when cooled in the air. The SEM pictures show that most of V exists in the crystalline phase, such as Na₁₆Ti₁₀O₂₈, NaAlSiO₄ and CaTiO₃, when quenched in air slowly; and the vanadium is dispersed in the slag phase when cooled by water quickly. Furthermore, the effects of additive NaOH on the reduction were also studied, results show that NaOH could enhance the separation of iron and slag, promote the transformation of vanadium and titanium and inhibit the vanadium enrichment in metal phase.

To achieve stable operation of coke oven batteries, a numerical model for estimating the clearance between the coke cake and the coke oven wall was developed. The influence of the temperature distribution in the combustion chamber on coke cake contraction was investigated by using the developed model. As a result, the following findings were obtained.

1) A temperature decrease on the coke side (CS) results in a larger clearance decrease at the end of the coke chamber compared with a temperature decrease on the machine side (MS).

2) The clearance decrease at the end of CS causes a higher pushing load of coke cake. Therefore, improvement of the CS flue gas temperature is particularly important for reducing the pushing force of the coke cake.

Reducibilities of silico-ferrite of calcium and aluminum (SFCA) have been examined using high temperature X-ray diffraction (XRD) analysis including the effect of hydrogen. Two types of sample were prepared: columnar SFCA covered with slag and acicular SFCA with fine pores, denoted by ‘Columnar SFCA’ and ‘Acicular SFCA’, respectively. These samples were synthesized using chemical reagents and iron ore powders. XRD analysis was applied to the samples heated in a condition simulating a blast furnace. Oxygen partial pressure was controlled by gas mixtures of CO–CO₂–He, and there were some cases where 3.9 vol% of hydrogen gas was added to the mixture. The microstructures of the samples before and after heating were observed by electron probe microanalysis (EPMA). XRD profiles indicated: (i) both SFCA samples were reduced to Fe via calcio-wüstite (CW), (ii) the reduction from ‘Acicular SFCA’ to CW took place at lower temperatures than that of ‘Columnar SFCA’ during the heating cycle up to 1000°C, and (iii) CW produced from ‘Acicular SFCA’ was reduced to Fe earlier than CW from ‘Columnar SFCA’ at 1000°C. These results suggest that SFCA and CW phases in ‘Acicular SFCA’ have higher reducibility than those in ‘Columnar SFCA’. This would be because ‘Acicular SFCA’ consists of smaller SFCA particles with fine pores. BSE images have revealed that such microstructures still remained even after the reduction to CW. Additions of hydrogen promoted reduction of ‘Columnar SFCA’ more effectively, and this effect was observed more remarkably in the reduction from CW to Fe than from SFCA to CW.

To investigate the migration behavior of MgO in oxidization-reduction of the vanadium titano-magnetite (VTM) ore and the influence of its content on the reduction, the solid solution was prepared by annealing MgO and Fe₃O₄ and its oxidized products were isothermally reduced under H₂–N₂ atmosphere. It was found that the sinter with high MgO content showed faster reduction rate due to the formation of pores and the increase of active sites caused by the diffusion of Mg²⁺ into Fe₃O₄. In contrast, the as-oxidized sinter with high MgO content showed the slowest reduction rate due to the formation of high content of (MgO)_x·(FeO)_1−x (x=0.239–0.77) in the reduction process. MgO migrated from the solid solution and combined with partial Fe₂O₃ to form MgFe₂O₄ in the oxidization stage. In the reduction stage, MgO migrated outward as the oxidized sinter was reduced to Fe₃O₄ and FeO, while migrated inwards in the stage of Fe formation. MgO content in per mole formed FeO decreased as the sinter was reduced to FeO, while increased in solid solution during the Fe generation stage.

To investigate the dominant factors that affect carbonization process of coal, behaviors of caking and low-quality coals (i.e., non- or slightly caking coal) during carbonization were examined. The release of volatile matter of pulverized coal samples in carbonization was evaluated using thermogravimetric analyzer, and the fluidity of coal particles was measured by the Gieseler plastometry method. Furthermore, a single coal particle was heated under a nitrogen atmosphere, and the images of the samples were acquired. From the images, swelling onset temperature, maximum swelling temperature, and maximum swelling ratio were evaluated. The carbonized coal particles were imaged using X-ray Computed Tomography (CT), and their internal structure was investigated. Although the release behavior of volatile matter, fluidity, and swelling of each coal were different according to kinds of coal, many of the parameters associated with those behaviors would correspond to coal rank. Combined with the carbonization behavior of coal and pore structure of carbonized coal, the amounts of thermoplastic components and volatile matter may affect the softening and swelling of coal particles.

The paper examined the changes in microstructure and inorganic elements in their true mineral forms of the coke samples from various hearth locations using X-ray diffraction (XRD) and scanning electron microscopy-energy dispersive spectroscopy (SEM-EDS) after the blow out and cool down of a 2800 m³ industrial blast furnace. The results illustrate that all deadman coke samples from fines to lumps were confirmed to be highly graphitized. Furthermore, the deadman coke was filled up with the accumulated KAlSiO₄ during its descent process and the blast furnace slag which consisted of Ca₂MgSi₂O₇–Ca₂Al₂SiO₇ system and Ca₂ZnSi₂O₇ phases. Besides the slag phases, the iron was also observed in the deadman coke soaked in the iron layer. Those cause that the mass of the deadman coke is about 1.62–2.82 times larger than that of the feed coke under the same conditions. Thus it may make the deadman which was designed to float sit on the hearth bottom as the permeation of the slag and the liquid iron into the deadman coke was not taken into consideration during the design process. We concluded that the slag phase below the taphole level is primarily derived from the blast furnace slag. Moreover, the deadman coke carrying final slag may come in contact with the hearth bottom and react with ceramic pad or carbon brick with a sitting deadman, thereby it results in degrading the hearth lining. Meanwhile, the slag phases below the taphole level can provide the material for the formation of skull to protect the hearth lining.

Gas absorption phenomenon during tapping from converter to ladle was studied by water model experiments in water-air system. Changes of dissolved oxygen, DO in the vessels were observed by DO meter and volumetric coefficient for gas absorption rate, Ak_O was estimated. In addition, the amount of gas engulfment at plunging pool with the same water model was also observed and calculated with CFD method. Observed and Experimental results were in good agreement with each other when the criterion for liquid ratio was set as 0.5. Surface area of bubbles at plunging pool calculated by CFD was used to estimate mass transfer coefficient, k_B. k_B in tapping and bottom bubbling experiments were discussed and linear relation between k_B and stirring power density was found. The Plunging pool was divided into 9 parts and velocity at bubble surface and area of bubbles in each part were calculated. It was found that the center and middle part of plunging pool was the main gas absorption region through this discussion. Bubble generation phenomenon at plunging pool in steel-air system was also calculated with the same mesh as water-air system and calculated results were compared with each other.

A series of laboratory-scale experiments were carried out in order to elucidate the reaction mechanism between high Mn-high Al steel and CaO–SiO₂-type molten mold flux at 1450°C, which represents the reaction taking place during continuous casting of the steel. Compared to the previous study [Kim et al., Metall. Mater. Trans. 44B (2013) 299–308], high Al content in the liquid steel ([pct Al]₀ = 5.2) and high MgO content in the liquid flux ((pct MgO)₀ = 5 to 15) were employed, in order to confirm change of rate-controlling step from mass transport of Al in liquid steel to more complicated steps including mass transport in liquid flux. It was found that Al₂O₃ was rapidly accumulated near the interface of the flux, and SiO₂ and Na₂O were reduced simultaneously, regardless of (pct MgO)₀. At the early stage of the reaction (1 min), MgAl₂O₄ particles were observed in the flux near the interface, then the particles were spreading out into the bulk flux as the reaction time passed. Other solid phases (CaAl₄O₇, Al₂O₃) were also observed due to local depletion of MgO in the flux. The MgAl₂O₄ formation mechanism and its effect on mass transfer in the molten flux were discussed. A series of simple kinetic analyses showed that the mass transport of Al in liquid steel is no more controlling the reaction rate. It was concluded that there were possibilities of mass transport in the flux phase contributing reaction rate controlling step.

In the steel industry, slabs are manufactured with different amounts of alloying elements according to production purposes or final products. Because slabs have similar shapes, product identification is required to prevent inadequate production processes. In many steel mills, paint marking systems are widely used to inscribe slab identification numbers (SINs). As smart factory technology receives more attention in recent years, automatic recognition of SINs becomes more important for factory automation. The recognition of SINs is a challenging problem due to complex background of factory scenes and low quality of characters in SINs. To address this difficulties, this paper proposes a deep learning algorithm for recognizing SINs in factory scenes. Most existing recognition algorithms conduct text detection and classification using separate modules, and errors in each step are accumulated. The proposed algorithm employs a fully convolutional network (FCN) with deconvolution layers to integrate the recognition processes and improve the performance in processing time and accuracy. The main contribution of this work is on a novel type of ground-truth data (GTD) for the training of a FCN to recognize SINs in factory scenes. The relation between an input image and the corresponding GTD is directly trained in the manner of image-to-image training, and the FCN generates a prediction map that contains categorical information of individual pixels in an input image. Experiments were thoroughly conducted on industrial data collected from a steelworks to demonstrate the effectiveness of the proposed algorithm.

Due to an increasing demand for automobile outer panels with sharper streamlines, surface roughening during press forming is recognized as an important problem to be solved. However, although sheets are subjected to various deformation modes during press forming, the influence of deformation mode on surface roughening is not yet understood. Moreover, surface roughening behavior in Interstitial Free (IF) steels, which are now commonly utilized for outer panels, has not been studied. In this study the effect of deformation mode on surface roughening behavior in IF steels was examined, focusing in detail on the effect of texture development. Differences in surface roughness development and changes of microstructure were examined under equi-biaxial and plane-strain tension, using a macroscopic Marciniak test and microscopic in-situ observations. In addition, the influence of the distribution of crystal orientations on surface roughness development was numerically examined using a crystal plasticity finite-element analysis. The results showed that surface roughening was larger for plane-strain tension than for equi-biaxial tension regardless of the IF steels tested, due to the larger difference in deformation resistance among crystal grains depending on crystal orientation. It is therefore suggested that surface quality after press forming could be improved by reducing the difference in deformation resistance among the grains.

Precise control of the sheet thickness in cold rolling process is becoming more and more important. Sheet thickness after cold rolling, depends on the initial thickness, the initial roll gap as well as the elastic deformation of the rolling mill including rolls. As production amount of high-strength materials increases, the elastic deformation of rolls becomes larger and more important. In cold rolling of high-strength steel sheets, it is reported that the rolls show non-circular deformation and the rolling load predicted by Hitchcock’s equation does not agree with experiments. In this study, stainless steel sheets were cold-rolled and the profiles of the partly rolled specimen were measured by a laser profilometer. Meanwhile, data processing method to reveal the sheet profiles in the roll bite is developed. Flattened radius R_P calculated from the measured profile is smaller than the radius R_H estimated by Hitchcock’s equation. In addition, the contact length L_M estimated with consideration of elastic deformation of the sheet is in good agreement with L_P measured by a laser profilometer under three lubrication conditions.

In the present study, a quality monitoring system in small scale resistance spot welding based on dynamic resistance and neural network was proposed. A precise dynamic resistance curve was obtained directly from the welding machine. Dynamic resistance and nugget development were found sensitive to the change of welding current. The second resistance peak gradually disappeared as welding current increased. Weld expulsion occurrence could be detected through the obvious fluctuation in dynamic electrical signals. The resistance valley and end resistance were found highly correlated with the nugget development. Features extracted from the dynamic resistance curve were selected as inputs in the neural network analysis. A reliable weld quality estimation through the neural network approach could be finally achieved.

The effects of fine-particle peening conditions on the surface-hardened layer properties of newly developed case-hardening steel, i.e., transformation-induced plasticity-aided steel with a chemical composition of 0.2% C, 1.5% Si, 1.5% Mn, 1.0% Cr, 0.2% Mo, and 0.05% Nb (mass%) were investigated for the fabrication of automotive drivetrain components. The surface roughness decreased with decreasing arc-height of fine-particle peening after vacuum carburization. A white layer developed on the surface of the steel peened at arc-heights greater than 0.41 mm (N). The maximum Vickers hardness and maximum compressive residual stress increased with increasing arc-height in the steel. These values were higher than those of commercial case-hardening steels. The increased volume fraction and expansion strain of the strain-induced martensite increased the hardness and compressive residual stress in the surface-hardened layer of the steel, although severe plastic deformation made a substantial contribution to enhancing the surface-hardened layer properties.

In order to obtain the low cost seamless pipe with the same mechanical properties of C–Mn–Mo steel, the application of controlled cooling technology was successfully applied to the C–Mn steel. Seamless pipeline C–Mn steel was prepared by on-line controlled cooling technique and corrosion behavior was studied experimentally by a salt spray wet/dry cyclic corrosion test. Seamless pipeline C–Mn–Mo steel by traditional quenching and tempering technique was used for comparison. Corrosion rate of C–Mn steel is higher than that of C–Mn–Mo steel in the whole test time. The rust layer of C–Mn steel contained more defects than that of C–Mn–Mo steel. Controlling cooling technique can significantly improve the strength of reduced-cost C–Mn steel, while the phenomenon can not reproduce for corrosion resistance.

For the development of advanced steels, reverse phase transformation from ferrite is essentially important to control the austenite phase in the heating process. Formation of austenite during the initial stage of reverse transformation from the recrystallized ferrite in low carbon steel has been studied from the view-point of the orientation relationships and the interphase boundary structure. At high temperature, the in situ electron backscattering diffraction (EBSD) measurement of austenite grain growth during the reverse transformation indicates that the different migration behaviors according to different α/γ interfaces, deriving from the interfacial coherency with the specific orientation relationships. The orientation and microstructure of the interface between ferrite and austenite have been investigated using the 3D crystal orientation analysis and transmission electron microscopy (TEM) observations. When the crystal orientation relationship between ferrite and austenite grain are close to the Kurdjumov–Sachs relationship, the grain boundary normal itself is also close to the {111}_γ and {011}_α, respectively. The microstructure of these interfacial planes is revealed to be flat using 3D-EBSD and TEM analysis. These coherent planes are strongly connected to the formation of the austenite phase on heating and also affect the slow migration of the grain-growth process.

The effect of strain rate on the hydrogen embrittlement property of an ultra high-strength TRIP-aided bainitic ferrite (TBF) steel with bainitic ferrite matrix was investigated to clarify the correlation between the transformation behavior of retained austenite and the hydrogen embrittlement fracture behavior of the TBF steel. Tensile tests were carried out at the strain rates between 5.56 × 10⁻⁶ and 2.78 × 10⁻²/s without and with hydrogen charging. Hydrogen analysis after tensile tests was conducted by using thermal desorption spectroscopy (TDS). Fracture strain decreased with decreasing the strain rate due to the hydrogen absorption to the TBF steel although fracture strain without hydrogen charging slightly increased with decreasing the strain rate. However, it was observed that transformation behavior of retained austenite was hardly changed by the hydrogen absorption and the change in the strain rate. When tensile test was carried out to the TBF steel at the slow strain rate with hydrogen charging, fracture surface of quasi cleavage fracture containing flat facet, which was fractured transformed martensite, was obtained and the crack perpendicular to the tensile direction was observed near transformed martensite. It was considered that the decrease in the resistance to hydrogen embrittlement of the TBF steel tensile tested at the slow strain rate was attributed to the initiation of flat facet and the hydrogen concentration at the crack tip due to the hydrogen diffusion from transformed martensite during tensile testing at slow strain rate.

Hydrogen entry behavior was investigated with different H₂S partial pressures over pH 5.0 and was quantified for various sour environments using American Petroleum Institute grade X65 line pipe and line pipe plate. Hydrogen permeability dramatically decreased for H₂S partial pressures of 0.1 MPa exceeding pH 5.5 and 0.01 MPa exceeding pH 6.2. This is caused by the formation of a stable iron sulfide film. On the other hand, hydrogen permeability proportionally decreased with increasing pH for H₂S partial pressure of 0.001 MPa up to pH 6.0. The critical pH at which iron sulfide becomes quite stable was predicted from the equation of the relation among pH, H₂S partial pressure, and iron ion activity based on potential vs. pH in Fe–S–H₂O. Hydrogen concentration into steel invading from various sour environments was proposed and quantified.

After final annealing, magnetic properties of the electrical steel including the lowest bulk sulfur content are best, while those of the steel including the additional aluminum is worst. Such excellent magnetic properties in the former steel is due to the strong cube + Goss texture. The sulfur highly segregated at grain boundaries and the surface Al₂O₃ layer is detrimental to the selective growth of the cube and Goss grains, resulting in the smaller grain size and the final poor magnetic properties. The bulk content of sulfur and aluminum should, therefore, be decreased to obtain the strong cube + Goss texture.

It is known that the micro-strain in cold worked iron can be evaluated by the classical Williamson-Hall method using the three data of diffraction peaks: {110}, {211} and {220}. It is not clarified that the obtained value gives the true micro-strain or not. In addition, the accuracy of analysis is not so high because the diffraction strength from {220} plane is generally very weak. In this paper, three methods, i.e. classical Williamson-Hall method, Diffraction Young’s Modulus Correction method and Direct Fitting method, ware attempted to reconfirm the reasonability of the classical Williamson-Hall method and to estimate accurate values of the parameter α and the micro-strain ε in the Williamson-Hall equation. The results obtained are as follows: 1) Elastic anisotropy in the Williamson-Hall plots is corrected using the parameter ω which relates to the values of diffraction Young’s modulus. 2) The optimal values of parameter ω can be determined by the Direct Fitting method, which can be used to determine the timely orientation-dependent diffraction Young’s modulus (E^*_hkl) in cold worked specimens. 3) It was confirmed that the classical Williamson-Hall method can generally give reliable values for the parameter α and the micro-strain ε. 4) No large difference is found for the values of micro-strain ε from the three methods. 5) There is a clear linearity between the micro-strain ε and yield stress in cold rolled iron specimens.

A new process for resource utilization of converter gas is proposed to produce CO₂ at a lower cost in this paper. Converter gas is burned in O₂–CO₂ atmosphere instead of O₂–N₂ atmosphere in a closed combustion furnace. Flue gas with high concentration of CO₂ can be used as low-grade CO₂ directly or be purified as raw gas for preparing high-purity CO₂. Both the low-grade CO₂ and high-purity CO₂ can be recycled for the converter blowing and sealing. In order to analyze the combustion characteristics of converter gas, numerical simulations based on a three dimensional combustion furnace model were carried out. Volume fraction of N₂ in combustion flue gas dropped from 63.37% to 3.92%, meanwhile CO₂ concentration in flue gas reached 95.08% when the N₂ in air was totally replaced by CO₂. In addition, to control the CO content in flue gas within the safety criterion, lower stoichiometry is an optimal solution for the converter gas compete combustion.