Permanent magnets, which are high-performance electric-motor components, contain rare-earth elements (REEs) that pose a high resource risk. REEs are present in machining scrap and end-of-life waste from these magnets. In CaO–SiO₂–FeO–P₂O₅ slags, elements concentrate in specific mineral phases. These slags are produced inexpensively from common fluxes and iron, which constitutes most of the magnet. Based on this enrichment behavior, this study focused on a “slag extraction method,” in which the target element is concentrated in a specific mineral phase of a CaO–SiO₂–FeO–P₂O₅ slag, and that phase is separated from the others to obtain an ore substitute. However, because the REE enrichment behavior in this slag remained unclear, we investigated samarium to clarify its partitioning behavior in the slag during isothermal holding at 1400 or 1500°C and enrichment behavior during slow cooling to 1200°C. The results showed that samarium partitioned between the 2CaO·SiO₂–3CaO·P₂O₅ solid-solution phase (the C₂S–C₃P phase) and liquid phase. The samarium enrichment in C₂S–C₃P increased as the P₂O₅ concentration in C₂S–C₃P and treatment temperature decreased. Furthermore, when the slag had a eutectic composition of the C₂S–C₃P and wüstite phases, samarium did not dissolve into the wüstite phase formed during slow cooling; therefore, samarium partitioned only into C₂S–C₃P.

The mechanical strength of metallurgical coke is strongly affected by its internal pore structure; however, the influence of three-dimensional pore network connectivity on stress concentration behavior remains unclear. In this study, the relationship between three-dimensional pore network structures and mechanical response in metallurgical coke was investigated by combining X-ray computed tomography (CT), pore network analysis, and finite element method (FEM)–based stress analysis. Three-dimensional coke models reconstructed from X-ray CT images were analyzed using the Burn method, breadth-first search, and the maximal ball method to quantify pore size, connectivity, and network structure. Uniaxial compression was simulated using voxel-based FEM to evaluate the maximum principal stress. In addition, matrix properties were examined by nanoindentation and X-ray diffraction.

The results showed that a coke sample with a uniformly connected and isotropic pore network exhibited a systematic decrease in maximum principal stress with increasing pore size and relatively small stress variability. In contrast, a coke sample with an inhomogeneous and anisotropic pore network, characterized by isolated pores and insufficient throat development, showed large scatter in maximum principal stress, even for pores with similar sizes or coordination numbers. These findings indicate that stress concentration behavior is not uniquely determined by a single structural parameter but is instead influenced by the overall connectivity and spatial arrangement of the three-dimensional pore network.

To achieve carbon-neutral steel production, the authors propose the Carbon Recycling Ironmaking Process using Deposited carbon-iron oxide Composite (CRIP-D), which recycles carbon throughout the ironmaking route without relying on fossil fuels. This study investigates the reduction and melting behavior of the deposited carbon–iron oxide composite (DCIC) under a hydrogen-reducing atmosphere.

Reduction experiments with increasing temperature revealed that DCIC reduction begins at approximately 420°C in the hydrogen atmosphere, significantly earlier than in inert conditions. The outlet gas composition was found to be predominantly governed by the equilibrium of the water gas shift reaction, resulting in a high carbon activity exceeding 1.0 above 550°C. This elevated carbon activity promoted both carbon deposition and carburization after reduced to metallic iron.

Carbon analysis showed that under the hydrogen atmosphere, reduced iron contained over 2% of carbon by 1150°C, while negligible carburization occurred under inert conditions. Iron nugget formation, indicating melting, was first observed between 1225°C and 1250°C, corresponding to the melting point of Fe₃C. XRD analysis demonstrated a decrease in Fe₃C peak intensity above 1250°C despite a constant average carbon content, suggesting redistribution of carbon from Fe₃C to austenitic iron.

These findings demonstrate that hydrogen-reducing conditions significantly enhance reduction, carburization, and melting of DCIC, supporting the feasibility of CRIP-D as a pathway to fossil-free ironmaking.

In the point of view of reducing coke production cost and future resource depletion, it is necessary to produce high-strength coke from low-rank coal.

It is reported that high strength coke can be obtained by pulverizing, compacting, and carbonizing low-rank coal, non- or slightly-caking coal. In this study, we research the effects of coal size and coal charging density on coke strength and coke density, and discuss the mechanism for the change of coke properties. Coal of 1.0 mm or less to 0.1 mm or less was compacted to 0.8 g/cm³ to 1.1 g/cm³, carbonized at 900°C, and coke strength and coke density were measured.

As a result, it was found that coke strength significantly increased by pulverizing to 0.1 mm or less and increasing the coal charging density. The effects of coal particle size and coal charging density on coke properties are examined. When the grain size of coal becomes finer, swelling is suppressed, and large pores and connecting pores of coke are reduced. As the coal charging density increased, the coke density increased due to the shortening of the distance between coal particles.

Argon bottom blowing stirring approach is used to ensure ladle refining performance, which is dependent on the flow field characteristics of molten steel. In this work, the flow characteristics of molten steel in ladle under various bottom blowing patterns were systematically investigated using physical simulation and numerical simulation. The results show that for dual porous plug ladle, the strong–weak plume system formed by different-flowrate bottom blowing effectively weakens the collision between adjacent plumes, which serves as the main reason for improving molten steel flow characteristics. Meanwhile, the asymmetric arrangement of porous plugs realizes staggered plume distribution and increases plume spacing to further reduce plume collision. The near-wall plume and far-wall plume reconstruct the overall plume structure, which synchronously enhances the molten steel flow at the ladle wall and central zone, and reduces the volume distribution of dead zones. For a 120-ton ladle, the bottom-blowing pattern featuring the total flow rate is 400 L·min⁻¹, and the flow rate ratio is 2:1 increases the total kinetic energy of molten steel by 25.93% and decreases the dead zone volume fraction by 30.42%, under the arrangement of two porous plugs of 0.67r-0.7r and 60°, compared with the pattern with 0.67r-0.67r and 1:1.

Ladle change (including ladle emptying and filling tundish) is a critical stage during steel continuous casting due to its direct impact on steel cleanliness and final steel quality. This study focuses on the ladle change operation in a four-strand bloom tundish applied for high-quality bearing steel production. The Volume of Fluid (VOF) model was used to systematically investigate the effects of ladle emptying time (2 min, 3 min) and filling flow rate (1.5 Q_steady, 2.0 Q_steady) on the steel-slag-gas flow, and on the mixing behavior of new and residual molten steel in this tundish. The results indicate that the formation of slag eye during the filling stage was initially caused by the bursting of rising bubbles from the ladle shroud at the steel‑slag interface, followed by its progressive expansion due to the sustained upward flow of molten steel. The filling flow rate exerted a greater influence on the slag eye area than ladle emptying time. When the filling flow rate was 2.0 Q_steady, the maximum slag eye areas were 568.3 cm² and 583.0 cm² for 2 min and 3 min emptying times, respectively. The emptying time was the key parameter governing steel mixing behavior. Increasing the filling flow rate shortened the mixing time, while extending the emptying time improved mixing efficiency. Based on these findings, a differentiated ladle change strategy was proposed for different steel grade requirements.

To overcome the limitations of traditional manual inspection and 2D (two-dimensional) image recognition in scrap identification, this paper proposes a novel method that fuses binocular vision and robust 3D Gaussian Splatting (3DGS) for accurate 3D scrap measurement. Fusion and scale recovery are achieved by leveraging the distance relationships between feature points in the preprocessed Structure from Motion (SFM) system of 3DGS and the binocular system, effectively resolving the scale drift issue and efficiency limitations inherent in traditional photogrammetry. Reconstruction experiments on seven scrap datasets show an average Root Mean Square Error (RMSE) of 0.82 mm, meeting the requirements for scrap measurement and classification. The analysis identifies the optimal fusion parameters as follows: a row alignment threshold of 2 pixels, a track length-based Non-Maximum Suppression (NMS) score, and a regression algorithm utilizing Trimmed-Mean (TM). Further evaluation through an ensemble learning model reveals that the robust regression algorithm contributes the most to reconstruction accuracy, followed by the row alignment threshold. Compared with traditional photogrammetry, our method achieves 5.4×faster processing on seven scrap datasets with minimal accuracy loss. Our method establishes a foundation for the application of 3D vision technology in scrap identification.

Accurate prediction of tapped steel weight is essential for improving process stability, alloy allocation, and production scheduling in basic oxygen furnace (BOF) steelmaking. However, the coexistence of high-frequency industrial disturbances, material carryover between heats, and poor-quality boundary data poses significant challenges to conventional prediction methods. To address these issues, this study proposes a Uncertainty-Aware Compensation Long Short-Term Memory (UAC-LSTM) framework that integrates sample-level heteroscedastic Gaussian modeling with a domain-knowledge-driven error compensation mechanism. The heteroscedastic module dynamically estimates sample-specific noise variance via negative log-likelihood (NLL) optimization, enabling adaptive weighting of noisy and incomplete observations. Simultaneously, the compensation module identifies high-risk samples through residual pattern recognition and Bayesian decision rules, providing structured post-prediction corrections. Using 8013 heats of real production data from a large steel plant, UAC-LSTM achieved a root mean square error (RMSE) of 2.53 t and a coverage rate of 93.5% within the ±3% error interval, outperforming state-of-the-art models including SVR, RF, XGBoost, DNN, GRU, and Seq2Seq. Robustness experiments demonstrated that the model maintained stable performance under up to 20% Gaussian noise injection, with RMSE only increasing to 2.61 t. Uncertainty analysis revealed a strong correlation between predicted variance and extreme errors, highlighting the interpretability and risk-awareness of the framework. This study offers a scalable, interpretable, and physically meaningful approach for BOF tapped steel weight prediction, providing practical insights for intelligent process monitoring and decision-making in complex industrial environments.

In the JIS standard case-hardening steel, spheroidizing annealing (SA) can improve cold forgeability but inevitably results in uneven distribution of carbides (a mixture of lamellar carbides and spherical carbides, and uneven carbides distribution). This degrades cold forgeability and promotes grain coarsening during carburizing. To address this problem, new SA methods combined with suitable chemical compositions are required to achieve a uniform distribution of carbides in a hypoeutectoid steel.

This study found that an ideal structure with uniformly dispersed carbides can be obtained by using high-Cr case-hardening steel with ferrite - pearlite as the initial microstructure and holding it below the A₁ temperature by precipitation of carbides in ferrite in addition to the dissolution of cementite in pearlite. Microstructure analysis and thermodynamic calculations revealed that in high-Cr compositions, the chemical potential of C in pearlite is significantly higher than in ferrite, which drives C diffusion from pearlite to ferrite, and precipitation of spherical carbides in ferrite. Furthermore, it was found that AlN particles in ferrite act as nucleation site of carbides, promoting precipitation of carbides in ferrite.

As the properties of metallic materials depend on their crystallographic texture, the development of a strong texture is essential for functional control. This study achieved a single-crystalline-like texture with a hierarchical microstructure in 316L stainless steel (316L SS) through laser-powder bed fusion (L-PBF). A conventional ±X-scan strategy (employing bi-directional scanning along the x-axis for all layers) afforded a unique microstructure in which two differently oriented grains (<001> and <011> along the z-axis) appeared alternately. On the other hand, a novel ±X-brick-scan strategy, which employs bi-directional scanning along the x-axis for all layers while shifting the laser irradiation position by half a pitch between layers, afforded a single-crystalline-like texture with <011> orientation along both the z- and y-axes. The difference in texture originates from the crystal growth direction. In the ±X-brick-scan strategy, the <001> growth direction was inclined by approximately ±45° from the z-axis, whereas it aligned parallel to the z-axis at the melt pool bottom in the ±X-scan strategy. Notably, this growth behavior does not fully correspond to the calculated heat-flow direction in the ±X-brick-scan case. These results suggest that the crystal growth direction in L-PBF is governed not only by heat flow, but also by additional factors such as interfacial energy.

One of the typical components of oxide inclusions in steel is Al₂O₃–SiO₂–MnO–MgO system. The equilibrium relations of Al₂O₃–SiO₂–MnO–MgO slag with Fe–Ni alloy at 1733 K was experimentally obtained, which is necessary for knowing the effects of MgO addition to the previously known Al₂O₃–SiO₂–MnO system. The experiments were conducted under various slag compositions mainly focusing on the changes of Al₂O₃ saturation line and spinel formation line equilibrated with liquidus phase. SEM-EDS analysis was utilized to directly observe the phase image and to analyze the liquidus phase composition. Liquidus phase was then plotted on to the Al₂O₃–SiO₂–MnO–MgO system from different angles.