To meet the requirement of low magnetic permeability, which, in turn, lowers the ferrite content of castings, of special interest is 316 stainless steel, whose low ferrite content renders it suitable also for nuclear power applications. Therefore, the effects of the composition and cooling rate of 316 stainless steel castings on the ferrite content are investigated. Three 316 stainless steel continuous casting samples with different compositions (primarily differing in the Ni content) are studied, i.e., low-alloy type (L-316), medium-alloy type (M-316), and high-alloy type (H-316). The austenite-forming element nickel of three different industrial samples is 10%, 12%, and 14%, respectively. The effect of the cooling rate on the ferrite content and precipitation phases of the high Ni content of the 316 stainless steel casting (H-316) is studied by remelting experiments and different methods of quenching of liquid steel. In both cases, the ferrite content and the precipitate phases in the microstructure are analyzed using SEM and EBSD. The results indicate that compositional changes within the 316 stainless steel range lead to changes in the solidification mode. In the L-316 casting, solidified by the FA mode (ferrite–austenite mode), ferrite precipitates first from the liquid phase, followed by the formation of austenite, and the ferrite content is 11.2%. In contrast, the ferrite content in the M-316 and H-316 castings, solidified by the AF mode (austenite–ferrite mode), is 2.88% and 2.45%, respectively. The effect of the solidification mode on the ferrite content is more obvious than that of the composition. The microstructure of the L-316 casting is mainly composed of the austenitic phase and the ferritic phase. The microstructure of the M-316 casting is composed of austenite, ferrite, and a small amount of sigma phase, with a small amount of ferrite transformed into the sigma phase. The microstructure of the H-316 casting is basically composed of austenite and the sigma phase, with the ferrite has been completely transformed into sigma phase. Changes in composition have a greater influence on the precipitate phases, while the solidification mode has a lesser impact. In the remelting experiments, the ferrite content in the H-316 ingot obtained through furnace cooling and air cooling is 1.49% and 1.94%, respectively, and the cooling rates are 0.1 °C/s and 3.5 °C/s, respectively. Under oil- and water-cooling conditions, with cooling rates of 11.5 °C/s and 25.1 °C/s, respectively, the ferrite content in the ingot is controlled to below 1%. The effect of the cooling rate on the precipitation phase of the H-316L ingot is that the amount of precipitated phase in the ingot decreases with an increase in cooling rate, but, when the cooling rate exceeds a certain value (air cooling 3.5 °C/s), the change in cooling rate has little effect on the amount of the precipitated phase.

Object detection in complex road scenes is vital for autonomous driving, facing challenges such as object occlusion, small target sizes, and irregularly shaped targets. To address these issues, this paper introduces DGSS-YOLOv8s, a model designed to enhance detection accuracy and high-FPS performance within the You Only Look Once version 8 small (YOLOv8s) framework. The key innovation lies in the synergistic integration of several architectural enhancements: the DCNv3_LKA_C2f module, leveraging Deformable Convolution v3 (DCNv3) and Large Kernel Attention (LKA) for better the capture of complex object shapes; an Optimized Feature Pyramid Network structure (Optimized-GFPN) for improved multi-scale feature fusion; the Detect_SA module, incorporating spatial Self-Attention (SA) at the detection head for broader context awareness; and an Inner-Shape Intersection over Union (IoU) loss function to improve bounding box regression accuracy. These components collectively target the aforementioned challenges in road environments. Evaluations on the Berkeley DeepDrive 100K (BDD100K) and Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) datasets demonstrate the model’s effectiveness. Compared to baseline YOLOv8s, DGSS-YOLOv8s achieves mean Average Precision (mAP)@50 improvements of 2.4% (BDD100K) and 4.6% (KITTI). Significant gains were observed for challenging categories, notably 87.3% mAP@50 for cyclists on KITTI, and small object detection (AP-small) improved by up to 9.7% on KITTI. Crucially, DGSS-YOLOv8s achieved high processing speeds suitable for autonomous driving, operating at 103.1 FPS (BDD100K) and 102.5 FPS (KITTI) on an NVIDIA GeForce RTX 4090 GPU. These results highlight that DGSS-YOLOv8s effectively balances enhanced detection accuracy for complex scenarios with high processing speed, demonstrating its potential for demanding autonomous driving applications.

The measurement of mixing time in a water model of soft-stirring steelmaking ladles is practically facing a problem of bad repeatability. This uncertainty severely affects both the understandings of transport phenomenon in ladles and the measurement accuracy. Scaled down by a ratio of 1:4, a water model based on an industrial 260-ton ladle is used. This paper studies the transport process paths and mixing time of salt solution tracers in the water model of eccentric gas-stirred ladles with a low gas flow rate. After a large number of repeated experiments, the different transport paths of the tracer and the error of the mixing time in each transport path are discussed and compared with the numerical simulation results. The results of a large number of repeated experiments on the water model show that there are five transport paths for the tracer in the ladle. The tracer of the first path is mainly transported by the left-side main circulation flow, which is identical to the numerical simulation results. The tracer of the second and third paths are also mainly transported by the left-side circulation flow, but bifurcations occur when the tracer in the middle area is transported downward. In the third path, the portion and intensity of the tracer transferring to the right side from the central region is higher than in the second path. The fourth path is that the tracer is transported downward from the left, middle, and right sides with a similar intensity at the same time. While the tracer in the fifth path is mainly transported on the right side, and the tracer forms a clockwise circulation flow on the right side. The mixing times from the first transport path to the fifth transport path are 158.3 s, 149.7 s, 171.7 s, 134 s and 95.7 s, respectively, among which the third transport path and the fifth transport path are the maximum and minimum values among all transport paths. The error between the mixing time and the averaged mixing time at each monitoring point in the five transport paths of the tracer is between −34.7% and 40.9%. Furthermore, the error of the averaged mixing time of each path and the path-based average value is between 5.5% and 32.6%.

Mixing time, as a key parameter for evaluating ladle refining efficiency, has long attracted extensive attention from researchers. In typical experimental studies, salt solution tracers are introduced into ladle water models to assess the degree of mixing within the ladle. Previous studies have demonstrated that the volume of tracer can significantly influence the measured mixing time. However, the gas flow rates employed in these studies are generally relatively high, whereas, in industrial operations, especially during final composition adjustments, lower gas flow rates are often applied. To systematically investigate the effect of the salt solution tracer volume on the mixing efficiency in a ladle water model under asymmetrical gas stirring with a low gas flow rate, a 1:3-scaled water model was developed based on a 130-ton industrial ladle. The mixing behaviors corresponding to different tracer volumes were comprehensively analyzed. The results indicate that the relationship between tracer volume and mixing time is non-monotonic. As the tracer volume increases, the mixing time first decreases and then increases, reaching a minimum at 185 mL. When the tracer volume was small, the dimensionless concentration curves at Monitoring Point 4 exhibited two distinct patterns: A parabolic profile, which was when the tracer initially moved through the left and central regions and then slowly crossed the gas plume to reach the monitoring point. A sinusoidal profile, which was when the tracer predominantly circulated along the right side of the ladle. When the tracer volume exceeded 277 mL, the concentration curves at Monitoring Point 4 consistently exhibited a sinusoidal pattern. Compared with moderate gas flow conditions (8.3 L/min), the peak concentration at Monitoring Point 3 was significantly lower under a low gas flow (2.3 L/min), and the overall mixing time was longer, indicating reduced mixing efficiency. Based on the findings, a recommended tracer volume range of 185–277 mL is proposed for low gas flow conditions (2.3 L/min) to achieve accurate and efficient mixing time measurements with minimal disturbance to the flow field. It was also observed that when the tracer concentration was relatively low, the mixing behavior throughout the ladle became more uniform.

This study investigates the distribution characteristics of residual ferrite in 304L austenitic stainless steel continuous casting slab and the impact of heat treatment processes on its content. Through optical microscopy (OM), thermodynamic calculation software (Thermo–Calc) and heat treatment experiments, it is found that the residual ferrite content along the thickness direction at the width center of the slab exhibits an “M”-shaped distribution—lowest at the edges (approximately 3%) and highest near the center (approximately 13%). Within the triangular zone of the slab, the residual ferrite content varies between 1.8% and 12.2%, with its average along the thickness direction also showing an “M”-shaped distribution; along the width direction, the average residual ferrite content is lower at the edge positions, while within the internal triangular zone, it ranges between 8% and 10%. The ferrite morphology changes significantly across solidification zones: elongated in the surface fine-grain zone, lath-like and skeletal in the columnar grain zone and network-like in the central equiaxed grain zone. Thermodynamic calculations indicate that the solidification mode of the 304L continuous casting slab follows the FA mode. Heat treatment experiments conducted across the entire slab thickness demonstrate effective reduction in residual ferrite content; the optimal reduction is achieved at 1250 °C with a 48 min hold followed by air cooling while preserving the original “M”-shaped distribution characteristic after treatment. Increasing the heat treatment temperature, prolonging the holding time and reducing the cooling rate all contribute to reducing residual ferrite content.

A novel right-sidewall gas blowing method is proposed to improve the flow behavior in a single-strand tundish. Despite advances in tundish flow control, the impact of slag layers and sidewall gas injection on flow dynamics and tracer transport remains underexplored. This study combines 1:3.57 scale water model experiments and Compuational Fluid Dynamics (CFD) simulations to investigate the effects of gas injection heights (50 mm and 100 mm) on flow structure, mixing efficiency, and slag layer interactions. Particle Image Velocimetry (PIV) and the stimulus-response method are used for quantitative validation. Results show that sidewall gas blowing suppresses short-circuit flow, increases average residence time by up to 37%, and reduces dead zone volume by up to 19%. The 50 mm blowing height induces stronger surface turbulence, while the 100 mm height improves flow uniformity. The presence of a slag layer significantly dampens surface fluctuations and alters vortex formation. These findings fill a critical research gap in tundish metallurgy and offer a practical reference for optimizing gas blowing strategies in industrial applications.