Explainable machine learning model for predicting molten steel temperature in the LF refining process

Accurate prediction of molten steel temperature in the ladle furnace(LF)refining process has an important influence on the quality of molten steel and the control of steelmaking cost.Extensive research on establishing models to predict molten steel temperature has been conducted.However,most researchers focus solely on improving the accuracy of the model,neglecting its explainability.The present study aims to develop a high-precision and explainable model with improved reliability and transparency.The eXtreme gradient boosting(XGBoost)and light gradient boosting machine(LGBM)were utilized,along with bayesian optimization and grey wolf optimiz-ation(GWO),to establish the prediction model.Different performance evaluation metrics and graphical representations were applied to compare the optimal XGBoost and LGBM models obtained through varying hyperparameter optimization methods with the other models.The findings indicated that the GWO-LGBM model outperformed other methods in predicting molten steel temperature,with a high pre-diction accuracy of 89.35%within the error range of±5°C.The model’s learning/decision process was revealed,and the influence degree of different variables on the molten steel temperature was clarified using the tree structure visualization and SHapley Additive exPlana-tions(SHAP)analysis.Consequently,the explainability of the optimal GWO-LGBM model was enhanced,providing reliable support for prediction results.

Study on Flow and Mixing Characteristics of Molten Steel in RH and RH-KTB Refining Processes

The flow and mixing characteristics of molten steel during the vacuum circulation refining, including RH(Ruhrstahl Heraeus) and RH KTB(Ruhrstahl Heraeus Kawasaki top blowing) processes, were investigated on a 1/5 linear scale water model of a 90 t multifunction RH degasser. The circulation rate was directly and more accurately determined, using a new method by which the more reliable results can be obtained. The fluid flow pattern and flow field in the ladle were demonstrated, observed and analyzed. The mixing time of liquid in the ladle was measured using electrical conductivity method. The residence time distribution in the RH model was obtained by tracer response technique. The influence of the main technological and geometric factors, including the gas top blowing (KTB) operation, was examined. The results indicated that the circulation rate of molten steel in the RH degasser can be fairly precisely calculated by the formula: Q lp =0.0333 Q 0.26 g D 0.69 u D 0.80 d(t/min), where Q g-the lifting gas flow rate (NL/min); D u and D d-the inner diameters of the up and down snorkels (cm), respectively. The maximum value of circulation rate of molten steel in the case of the 30 cm diameters either of the up and down snorkels for the RH degasser (the “saturated” rate) is approximately 31 t/min. The corresponding gas flow rate is 900 NL/min. Blowing gas into the vacuum chamber through the top lance like KTB operation does not markedly influence the circulatory flow and mixing characteristics of the RH process under the conditions of the present work. There exist a major loop and a large number of small vortices and eddies in the ladle during the RH refining process. A liquid liquid two phase flow is formed between the descending stream from the down snorkel and the liquid around the stream. All of these flow situation and pattern will strongly influence and determine the mixing and mass transfer in the ladle during the refining. The correlation between the mixing time and the stirring energy density is τ m∝ε -0.50 for the RH degasser. The mixing time rapidly shortens with an increase in the lifting gas flowrate. At a same gas flow rate, the mixing times with the up and down snorkel diameters either of 6 and 7 cm are essentially same. The 30 cm diameters either of the up and down snorkels for the RH degasser would be reasonable. The concentration time curve showed that three circulation cycles are at least needed for complete mixing of the liquid steel in the RH degasser.

Physical Modeling of the Vacuum Circulation Refining Process of Molten Steel

The available studies in the literature on physical modeling of the vacuum circulation (RH, i.e. Ruhrstahl Heraeus) refining process of molten steel have briefly been reviewed. The latest advances made by the author with his research group have been summarized. Water modeling was employed to investigate the flow and mixing characteristics of molten steel under the RH and RH KTB (Kawasaki top blowing) conditions and the mass transfer features between molten steel and powder particles in the RH PTB (powder top blowing) refining. The geometric similarity ratio between the model and its prototype (a multifunction RH degasser of 90 t capacity) was 1:5. The effects of the related technological and structural factors were considered. These latest studies have revealed the flow and mixing characteristics of molten steel and the mass transfer features between molten steel and powder particles in these processes, and have provided a better understanding of the refining processes of molten steel.

Vacuum Treatment for Simultaneous Desulphurization and Dephosphorization of Hot Metal and Molten Steel

The vacuum treatment for simultaneous desulphurization and dephosphorization of hot metal and molten steel with pre-melted CaO-based slag was carried out.For pre-treatment of hot metal,both desulphurization and dephosphorization are improved with the increase of CaO in slag,but deteriorated with the increase of CaF2 in slag.The average desulphurization and dephosphorization rate is 68.83 % and 78.46 %,respectively.For molten steel,the substitution of BaO for CaO in slag has minor effect on simultaneous desulphurization and dephosphorization.The desulphurization and dephosphorization rate is higher than 90% and 50% respectively with the lowest final sulfur and phosphorus mass percent being 0.001 2% and 0.010%,respectively.The overall effect of simultaneous desulphurization and dephosphorization of molten steel is better than that of hot metal.

Change and control of nitrogen in molten steel production process

The change and control of nitrogen content in molten steel was investigated through the production process of "LDBAr-LF-RH-CC". Results show that nitrogen content reduces gradually in converter-steelmaking stage, rises rapidly from the end of converter process to the end of argon station process, continues to increase in ladle furnace process, and decreases slightly in RH refining stage. Since nitrogen is removed mainly by BOF steelmaking and vacuum refining operations, nitrogen in molten steel should be removed as much as possible in these two operations. However, nitrogen uptake should be minimized in other operations of molten steel production process.

Copper Removal from Molten Steel by Gasification

A novel method is proposed to enhance the gasification and removal of copper from molten steel by adding ammonium salts or urea into molten steel under normal pressure.The decopperization experiments were conducted in a molybdenum-wire resistance furnace at 1 873 K.The copper content of about 400 g of a mild steel was reduced from 0.49%(mass fraction,the same below) and 0.51% to 0.31% and 0.38% using 0.7 g of NHC1 and 0.5 g of(NH)COrespectively,while the copper content of the molten steel was reduced from 0.61%to 0.56% using 2.00 g of NHCONH.

Effects of oxygen potential and flux composition on dephosphorization and rephosphorization of molten steel

The dephosphorization experiments of low phosphorus containing steel by CaO-based and BaO-based fluxes were carried out. The effects of the oxygen potential in molten steel and the BaO content in the slag on dephosphorization and rephosphorization of molten steel were analyzed. The results showed that the dephosphorization ratio of more than 50% and the ultra-low phosphorus content of less than 0.005% in steel were obtained by the three kinds of dephosphorization fluxes as the oxygen potential of molten steel higher than 400×10^-6. Rephosphorization of molten steel was serious as the oxygen content of molten steel lower than 10×10^-6. BaO-based fluxes can improve the dephosphorization effect and reduce the phosphorus pick-up effectively under the condition of weak deoxidization of molten steel （the oxygen potential is about 100×10^-6）, but can not prevent rephosphorization under the condition of deep deoxidization of molten steel （the oxygen potential less than 10×10^-6）.

Wettability of Molten Steel/Na2O-Li2O-SiO2-B2O3 Slag System

The vacuum treatment for simultaneous desulphurization and dephosphorization of hot metal and molten steel with pre melted CaO based slag was carried out. For pre treatment of hot metal, both desulphurization and dephosphorization are improved with the increase of CaO in slag, but deteriorated with the increase of CaF 2 in slag. The average desulphurization and dephosphorization rate is 68 83 % and 78 46 %, respectively. For molten steel, the substitution of BaO for CaO in slag has minor effect on simultaneous desulphurization and dephosphorization. The desulphurization and dephosphorization rate is higher than 90 % and 50 % respectively with the lowest final sulfur and phosphorus mass percent being 0 001 2 % and 0 010 %, respectively. The overall effect of simultaneous desulphurization and dephosphorization of molten steel is better than that of hot metal.

Nitrogen Removal from Molten Steel under Argon DC Glow Plasma

Under argon DC glow plasma, the nitrogen removal from molten steel was studied. The experimental result showed that nitrogen mass percent could be reduced to 0000 8%. The change of polarity had no impact on nitrogen removal when the nitrogen mass percent was low. The mechanism of denitrogenation of molten steel under argon DC glow plasma was discussed.

Copper Elimination from the Molten Steel by Addition of Hydronitrogens

A novel method has been proposed to remove copper from molten steel by adding the compounds of hydrogen and nitrogen into the melt at normal pressure. Feasibility experiments were carried out in Mo-wire resistance furnace at 1600℃. The copper content of about 400 g 45 steel were reduced from 0.60% and 0.57% to 0.41% and 0.51% with 3.25g and 1.20 g NH_4Cl respectively, while the copper contents of about 300 g steel melt were reduced from 1.15% and 0.61% to 0.90% and 0.56% with 4.10 g and 2.00 g NH_2CONH_2. These results indicate that the proposed method is very promising to be put into practical and worthwhile of further study.

Measurement and AFPS control of molten steel level in strip casting

The measurement and control of the molten steel level are studied, which affect the quality of strip surface in strip casting. A system of molten steel measurement with a CCD (Charge Coupled Devices) sensor is designed, real-time measured data are given and its precision is analyzed. The level fluctuation model is derived, and an adaptive fuzzy-PID controller with supervisory control (AFPS) is proposed. The stability of the system is proved using Lyapunov theorem, and the simulation results are given when the model, the casting speed and the roll gap change. It is suggested that this kind of coupled nonlinear and time varying system is stable and robust using the designed AFPS controller.

Bubbling to Jetting Transition during Argon Injection in Molten Steel

Bubbling to Jetting Transition is of the outmost importance in metallurgical processes given that the flow regime influences the refining rates, the refractory erosion, and the blockage of injection nozzles. Bubbling to jetting transition during subsonic bottom injection of argon in molten steel is studied here. The effect of the molten steel height, the injection velocity, the nozzle diameter, and the molten steel viscosity on the jet height and the bubbling to jetting transition is numerically analyzed using Computational Fluid Dynamics. Five subsonic argon injection velocities are considered: 5, 25, 50, 100 and 150 m/s. Three values of the metal height are taken into account, namely 1.5 m, 2 m and 2.5 m. Besides, three values of the nozzle diameters are considered: 0.001 m, 0.005 m and 0.01 m. Finally, three values of the molten steel viscosity are supposed: 0.0067, 0.1 and 1 kg/(m·s). It is observed that for the argon-molten steel system, the bubbling to jetting transition occurs for an injection velocity less than 25 m/s and that for the range of viscosities considered, the molten steel viscosity does not exert significant influence on the jet height and the bubbling to jetting transition. Due to the jet instability at subsonic velocities, a second transition, namely jetting to bubbling, is appreciated.

Lance Design for Argon Bubbling in Molten Steel

Three lance designs for argon bubbling in molten steel are presented. Bottom bubbling is considered too. Geometries considered are straight-shaped, T-shaped, and disk-shaped. The bubbling behavior of these lances is analyzed using Computational Fluid Dynamics, so transient three dimensional, isothermal, two-phase, numerical simulations were carried out. Using the numerical results, the bubble distribution and the open eye area are analyzed for the considered lance geometries. The plume volume is calculated from the open eye area and the lance immersion depth using geometrical considerations. Among the three lance designs considered, disk-shaped lance has the bigger plume volume and the smaller mixing time. As the injection lance is deeper immersed, the power stirring is increased and the mixing time is decreased.

Velocity Monitoring of Molten Steel in a Continuous Casting Mold Using Three Submerged Entry Nozzle Designs

The horizontal and vertical velocity components of molten steel in a slab continuous casting mold produced by three different two-port Submerged Entry Nozzle (SEN) designs are monitored and compared using Computational Fluid Dynamics (CFD) simulations. These two ports designs correspond to a conventional cylindrical SEN, a plate SEN and an anchor-shaped SEN. Four monitoring points at the molten steel in the centered vertical plane were selected to track the horizontal and the vertical component of the velocity vector. Two of them are located near the free surface and the remaining two are located in the vicinity of the SEN discharge nozzles. Some statistical values of the time series of above the velocity components are analyzed and correlated with the Kelvin-Helmholtz instability and the Karman vortex streets, which cause mold powder entrapment in the molten steel.

Mechanism and simulation of droplet coalescence in molten steel

Droplet coalescence in liquid steel was carefully investigated through observations of the distribution pattern of inclusions in solidified steel samples. The process of droplet coalescence was slow, and the critical Weber number(We) was used to evaluate the coalescence or separation of droplets. The relationship between the collision parameter and the critical We indicated whether slow coalescence or bouncing of droplets occurred. The critical We was 5.5, which means that the droplets gradually coalesce when We ≤ 5.5, whereas they bounce when We > 5.5. For the carbonate wire feeding into liquid steel, a mathematical model implementing a combined computational fluid dynamics(CFD)–discrete element method(DEM) approach was developed to simulate the movement and coalescence of variably sized droplets in a bottom-argon-blowing ladle. In the CFD model, the flow field was solved on the premise that the fluid was a continuous medium. Meanwhile, the droplets were dispersed in the DEM model, and the coalescence criterion of the particles was added to simulate the collision-coalescence process of the particles. The numerical simulation results and observations of inclusion coalescence in steel samples are consistent.

A Feedforward Controller to Regulate the Chemical Composition of Molten Steel in a Continuous Casting Tundish

A feedforward controller for the automatic regulation of chemical composition of molten steel in the tundish of a continuous casting machine is proposed in this work. The flow of molten steel inside the tundish is modeled as a distributed parameter system, and the resulting partial differential equation is transformed into a set of ordinary differential equations by means of the finite differences technique. From the above set and using a proper boundary condition, a feedforward control law is synthesized. No experimental tests are reported, however, the dynamic performance of the controller is illustrated by means of numerical simulations.

Effect of nano-carbon black content on wetting phenomenon of molten steel and alumina-carbon ceramic filter substrates

The effect of nano-carbon black content(O,8 and 12 wt.%)on the wettability of molten steel on Al_(2)O_(3)-C substrates was investigated by the sessile drop wetting method at 1500℃ under argon atmosphere.At the beginning of the wetting experiment,the contact angle decreased with the increase in nano-carbon black content.As the wetting experiment progressed,FeAl_(2)0_(4) layer and sheet Al_(2)O_(3) layer were found at the interface between the molten steel and the Al_(2)O_(3)-C substrates with O and 8 wt.% nano-carbon black content,and the contact angle deceased with time.When the content of nano-carbon black was 12 wt.%,a large number of nano-Al_(2)O_(3) whiskers were observed,which made the contact angle between the molten steel and Al_(2)O_(3)-C substrate become large.Based on the scanning electron microscope and energy dispersive spectrometry results,the formation mechanism of FeAl2O4 layer and Al_(2)O_(3) layer and the interfacial reaction mechanism were proposed.

Purification performance on molten steel of novel Al_(2)O_(3)-based ceramic filter prepared from microporous powder and nano-Al_(2)O_(3) powder

Reticulated ceramic foam filters provide an effective way to purify molten steel by removing non-metallic inclusions.We proposed a novel strategy to improve the purification performance of Al_(2)O_(3)-based ceramic filters by using microporous corundum-spinel raw materials to replace dense raw materials.Three kinds of Al_(2)O_(3)-based ceramic filters fabricated from dense α-Al_(2)O_(3) micro-powder or microporous corundum-spinel powder were selected to carry out the immersion tests with molten steel.On the one hand,the higher surface roughness of the filter skeleton prepared from microporous raw materials increased the adsorption capacity of skeleton surface on inclusions in molten steel.On the other hand,the higher apparent porosity and larger pore size of the filter skeleton were more beneficial to the penetration of molten steel in the micropores of skeleton.The reaction process at the solid-liquid interface also improved the wettability of the interface between skeleton and molten steel,resulting in a larger penetration depth and a better adsorption effect on the inclusions.In summary,the novel Al_(2)O_(3)-based ceramic filter prepared with microporous corundum-spinel powder and addition of 5 wt.% nano-Al_(2)O_(3) powder reduced the total oxygen content of the steel from 40.2×10^(-4) to 12.7×10^(-4) wt.% by 68.4% and the Al content from 0.46 to 0.18 wt.% by 60.9% after immersion test,presenting the most excellent purification performance on molten steel.

Hybrid Model of Molten Steel Temperature Prediction Based on Ladle Heat Status and Artificial Neural Network

Aiming at the characteristics of the practical steelmaking process, a hybrid model based on ladle heat sta- tus and artificial neural network has been proposed to predict molten steel temperature. The hybrid model could over- come the difficulty of accurate prediction using a single mathematical model, and solve the problem of lacking the consideration of the influence of ladle heat status on the steel temperature in an intelligent model. By using the hybrid model method, forward and backward prediction models for molten steel temperature in steelmaking process are es- tablished and are used in a steelmaking plant. The forward model, starting from the end-point of BOF, predicts the temperature in argon-blowing station, starting temperature in LF, end temperature in LF and tundish temperature forwards, with the production process evolving. The backward model, starting from the required tundish tempera- ture, calculates target end temperature in LF, target starting temperature in LF, target temperature in argon-blo- wiag station and target BOF end-point temperature backwards. Actual application results show that the models have better prediction accuracy and are satisfying for the process of practical production.

Molten Steel Breakout Prediction Based on Thermal Friction Measurement

By measuring the pressure and piston stroke of hydraulic servo cylinders in the oscillation system of a caster mould,friction between mould and shell can be determined.The condition of solidified shell can be monitored by thermocouples fixed to the mould.Breakouts,whether caused by shell cracking,shell sticking or entrained impurities,can therefore be predicted and effectively avoided through monitoring the change characteristics of friction and thermocouples temperature during the process.