The control of MnS inclusions in steel is a system engineering which involves multiple processes including smelting, solidification, heating and rolling, and requires multi-variable cooperative control. In this paper, the research status of MnS inclusions at home and abroad was discussed in detail, and the influence mechanism of different elements, heat treatment and rolling process on MnS inclusions were expounded. In addition, the current process control measures of MnS inclusions were summarized. The morphology, size and quantity of MnS inclusions can be effectively controlled by adopting appropriate deoxygenation process, improving the oxygen content at the solidification front, reducing the basicity of refining slag, adopting appropriate Ca treatment process, adopting low basicity tundish covering agent, increasing the secondary cooling water volume, strengthening cooling and reducing electromagnetic stirring intensity. In recent years, some new ideas have been proposed to control the deformation treatment and diffusion distribution of MnS inclusions. The morphology, size and quantity of MnS inclusions in steel can be effectively controlled by adding Mg, Ca-Mg, Zr, Ce and Te elements into molten steel.

In order to reduce the content of harmful sulfur elements in steel, KR desulfurization method is widely used in steel enterprises to produce low sulfur steel, and the influence of desulfurizer on the desulfurization efficiency is especially critical. In this paper, the effect of components on liquidus temperature, sulfur capacity, desulfurizer consumption, melting point, viscosity and desulfurization rate of CaO-Al₂O₃-SiO₂-CaF₂-MgO slag system were analyzed, based on simplex lattice method, FactSage 8.1 thermodynamic software, hemispherical melting point meter, rotating viscometer and slag-metal experiment. It was concluded that, the liquidus temperature of desulfurizer increased with CaO addition, and decreased with CaF₂ addition, while the sulfur capacity of desulfurizer increased with CaO and decreased with Al₂O₃ and CaF₂ enhancement. Meanwhile, with CaF₂ and Al₂O₃ addition, the desulfurizer consumption and melting point changed little in slag with low content of SiO₂, while the consumption increased greatly and the melting point decreased obviously in slag with high content of SiO₂. The desulfurization efficiency of desulfurizer with melting point of 1 400-1 424 ℃, w(SiO₂) of 0-10 %, w(Al₂O₃) and w(CaF₂) of 5 % -15 %, w(MgO) of 6 % was better than that of desulfurizer used in an enterprise, and the comprehensive desulfurization ability of CaF₂was stronger than that of Al₂O₃. Combined with the high temperature viscosity, the desulfurization efficiency and the subsequent slagging-off process, No.2 desulfurizer was the best choice for KR desulfurization.

Based on the actual production data of 260 t converter, prediction models of the main components of endpoint slag,CaO, SiO₂, TFe and MgO, were established through four algorithms of machine learning algorithm,XGBoost (eXtreme Gradient Boosting), elastic regression, linear regression and AdaBoost (Adaptive Boosting). By optimizing the parameters, the determination coefficients R² of XGBoost endpoint slag composition prediction model is all above 0.8.The slag splashing time model was modeled using five algorithms: SVR (Support Vector Regression), LGBM (Light Gradient Boosting Machine), GBDT (Gradient Boosting Decision Tree), RF (Random Forest) and XGBoost, respectively. Then, the integrated slag splashing time model was obtained by integrating SVR, XGBoost and GBDT. Stacking integrated slag splashing time model improves the prediction effect of each single model, and the prediction hit rate is 89.95 % in the error range of ±20 s.

The establishment of accurate BOF endpoint prediction model is particularly important for the improvement of production efficiency and liquid steel cleanliness. This paper took the converter of Shougang Jingtang duplex process in converter process as the research object, Pearson correlation analysis was conducted on historical production data, and 15 independent variables most relevant to the endpoint temperature and carbon content of the converter were obtained. Three machine learning algorithms of BP neural network, limit learning machine (ELM) and support vector machine (SVM) were used to build the prediction model of converter endpoint. Then 160 sets of new sample data were selected to verify the prediction accuracy of the three models. The results show that the accuracy of the prediction model of the end-point temperature and carbon content of the converter under the SVM model is higher. The hit rate of the prediction error of the endpoint temperature within ±15 ℃ is 90.6 %, and the hit rate of the prediction error of the endpoint carbon mass fraction within ±0.01 % is 93.8 %. In addition, the prediction model of converter end point based on support vector machine algorithm shows that the hit rate of endpoint temperature within ±15 ℃ and carbon mass fraction within ±0.01 % of duplex process in converter process is 9.1 % and 14.4 % higher than that of the conventional process.

Aiming at the problem that the method of visual analysis of fluid motion in tundish by RTD curve combined with water model flow field is complicated, and the distribution of dead zone cannot be accurately given. In this paper, the analysis method of fluid motion pattern model in continuous casting tundish was proposed. Based on the analysis results of RTD curve obtained by physical simulation, the dead zone was determined by numerical simulation, and the relationship between the critical velocity of dead zone and the inlet velocity was determined. Based on the critical velocity of the dead zone, the proportion and regional distribution of the dead zone can be determined in real time and online by numerical simulation, and then the fluid motion patterns in the single-strand and multi-strand tundish can be evaluated. The influence of casting speed and tundish structure on the proportion and regional distribution of dead zone in tundish was investigated. The results show that the fluid motion morphology model can quickly determine the dead zone of tundish and realize the quantification and visualization of dead zone. At the same time, compared with the same single-strand tundish, the critical velocity of dead zone increases with the increase of casting speed, while the volume fraction of dead zone is opposite, and the distribution pattern of dead zone does not change. In addition, the size and regional distribution of the dead zone under different baffle hole diameters and elevation angles in the multi-strand tundish can be intuitively compared. When the hole diameter is 70 mm and the elevation angle is 19°, the volume fraction of dead zone in the multi-strand tundish is only 7.64 %, and the dead zone is concentrated near the central nozzle and the side wall.

Taking the twin-roll thin strip continuous casting and rolling molten pool with an inverted H-shaped arc distributor as the research object, a three-dimensional numerical calculation model with a width of 1 350 mm and a roll diameter of 500 mm was established. The flow state and temperature distribution were carried out to verify the rationality of the shape design of the inverted H-shaped outlet, and the influence of the same molten steel temperature and different casting speeds on the molten pool flow field and temperature field was analyzed. The results show that the inverted H-shaped outlet flow distributor performs well in the temperature distribution, flow state and solidification of the molten pool. When the temperature of molten steel at the inlet is 1 770 K, the casting speed is increased from 20 m/min to 40 m/min and 60 m/min, the liquid mass fraction at the outlet is changed from 0 to 0.15 and 0.47, the solidified shell is also reduced to 0.5 mm and 0.1 mm.The rational working casting speed is about 40 m/min.

In view of the instability of round billet quality during continuous casting of a fivemachine fivestrand caster, especially the problem that the oxidizing stuffing sand continuously pollutes the molten steel and affects the cleanliness of molten steel during the ladle change process, the influence of the stuffing sand on the quality of molten steel was studied with the five-strand tundish as the research object through thermodynamic calculation and water simulation experiments, and the optimization scheme to improve the flow behavior of the stuffing sand was proposed. The results show that the stuffing sand during the open casting process increases the oxygen content of molten steel and reduces the cleanliness of molten steel. During the unsteady casting process, slag entrapment occurs when the stuffing sand enters into the molten steel. The removal rate of the stuffing sand is related to the insertion depth of the ladle shroud, the flow rate of the ladle and the liquid level of the tundish. When the flow rate of the ladle is 3 m³/h, the immersion depth of the ladle shroud is 160 mm, and the liquid level of the tundish is 400 mm, the removal rate of the stuffing sand is 94 % at the highest.The optimized flow control device improves the flow behavior of the stuffing sand, and its removal rate reaches 99.5 % at the maximum. In addition, during steady casting process,the divergence of the response time and the average residence time of each flow is small, and the consistency of each flow is improved, which is conducive to uniform composition and temperature of molten steel.

In view of the problem of large grain NbC grain boundary precipitation and proeutectoid ferrite film induced continuous casting slab cracks in Nb containing low alloy steel, the solidification heat transfer model was established by using Matlab and the target temperature of each secondary cooling section was optimized. The secondary cooling water volume of continuous casting was improved by back calculation, which effectively increased the cooling rate of slab edges within the NbC precipitation temperature range and the slab temperature before straightening, and enhanced the hot plasticity of narrow edges of the slab before straightening. The research shows that the NbC in Q345 steel is basically completely precipitated in the upper temperature range of A_e3 (1 108～905 ℃). Under the original secondary cooling system of continuous casting, the cooling rate of the slab in the NbC precipitation range is low (4.33 ℃/s), which makes it difficult to avoid the precipitation of large particles of NbC at the austenite grain boundary. In addition, the temperature of the narrow edge of the continuous casting slab before straightening is lower than the temperature of the transformation A_e3, resulting in the formation of proeutectoid ferrite film at the austenite grain boundary, both of them can induce the formation of narrow edge cracks. Through the adjustment and calculation of the target temperature of the secondary cooling section 3-7 and the edge water volume, the narrow edge cooling rate of the slab is achieved to 5.92 ℃/s, which is conducive to the fine dispersion and precipitation of NbC. At the same time, the temperature before the narrow edge straightening of the slab is higher than A_e3 temperature, which avoids the formation of the proeutectoid ferrite film, can significantly improve the thermoplasticity of the Nb containing low alloy steel, and provides a theoretical reference for controlling the edge defects of the Nb containing low alloy steel under the CSP process.

In view of the distribution characteristics of inclusions in aluminum-killed 25Mn steel round billets, the two-dimensional morphology, compositions and quantity density of inclusions in steel at different radial positions were analyzed with SEM-EDS equipped with an automatic inclusion analyzer. The inclusions in the steel were extracted by non-aqueous electrolysis and their three-dimensional morphologies and compositions were observed. The calculation with FactSage8.0 software was also conducted. The results show that with the decrease of temperature in the solidification and cooling process, the Al₂O₃ content in inclusions increases gradually, the CaS content increases first and then decreases, and the CaO and SiO₂ contents decrease gradually. With the increase of [O] content in steel, the stable region of the single liquid steel phase decreases gradually, and the stable region of liquid oxide phase increases. The inclusions in steel can be mainly divided into type S containing SiO₂, type A containing Al₂O₃, type M containing MgO and type C containing CaO or CaS. With closing to the center of the round billet, the total number density of inclusions increases slightly, the number densities of types S, A and M inclusions increase slightly, and the number density of type C inclusions decreases slightly. With closing to the surface of the round billet, the compositions of CaO-Al₂O₃-MgO inclusions are located near the liquid phase area. With approaching the center of the round billet, the compositions of CaO-Al₂O₃-MgO inclusions are gradually away from the liquid phase area. The average compositions of these inclusions are gradually transferred from the liquid phase area to the solid phase area, and the MgO content in the inclusions increases. The types of inclusions extracted with non-aqueous electrolysis are basically consistent with those observed with inclusion automatic analysis apparatus.

In order to compare the removal rates of liquid non-metallic inclusions and solid nonmetallicinclusions, the removal rates of four typical inclusions Al₂O₃，MgO-Al₂O₃，CaO-Al₂O₃-CaS and CaO-Al₂O₃ in pipeline steel during RH vacuum treatment were compared. The results show that solid inclusions in molten steel are easier to remove than liquid inclusions. In order to find out the reason why liquid inclusions are difficult to remove, the behavior of CaO-Al₂O₃ inclusions in molten steel at 1 600 ℃ was in-situ observed by high temperature confocal scanning laser microscope. The results show that this kind of inclusion is not easy to polymerize and grow up. With the decrease of temperature, the size of the inclusion further increases, and then it is captured by the solidified matrix. The CaO-Al₂O₃ type inclusions≥20 μm in casting slab can cause the large size inclusions exceed the standard in the hot rolled plate.

At present, the gasification dephosphorization in the process of slag splash protection is an effective slag dephosphorization technology. In order to ensure the effect of smelting and recycling after converter slag gasification dephosphorization, an experimental study on the use of gasification dephosphorization slag as a return material for slagging and dephosphorization was carried out in the laboratory. The research results show that the gasification dephosphorization slag is used for the dephosphorization of molten iron with strong dephosphorization ability in the early stage and low dephosphorization rate at the end point. The dephosphorization ratio and dephosphorization rate of hot metal at the end point are 53.3 % and 0.16 %/min respectively. Comparing the preparation of dephosphorization agent, it can be seen that the dephosphorization effect of the dephosphorization agent is poor in the early stage, and the dephosphorization rate at the end point is high, the final hot metal dephosphorization ratio and dephosphorization rate are 91.6 % and 0.32 %/min, respectively. According to the dephosphorization advantages of the two dephosphorization agents, the mixed ratio of molten iron is used for dephosphorization. When a large proportion of gasification dephosphorization slag is used for dephosphorization of molten iron, the phenomenon of phosphorus reversion occurs; when the mixing ratio is 1∶4, the dephosphorization effect is the best, and the end point dephosphorization ratio is 64.4 %. The CO₂ emission reduction assessment was carried out for molten iron dephosphorization with a ratio of 1∶4 mixed slag  using the life cycle assessment method, and the CO₂ emission reduction per ton of steel was estimated from the beginning to the end of the system boundary by 6.034-10.34 kg, and the ton of steel could save 1.8-3.0 yuan of lime.