株洲中车时代电氣股份有限公司

Document-level relation extraction (DocRE) aims to identify the relations between entities in a document, which serves as a fundamental task in natural language processing. In DocRE, there inherently exists strong prior knowledge, for example, individuals and organizations tend to exhibit a "membership" relation rather than a "separation" relation. Leveraging this knowledge can effectively confine the model's prediction space and highlight potential relations between entities. However, existing DocRE works primarily focus on designing sophisticated models to implicitly encode document features, inadvertently neglecting this informative prior knowledge, which might lead to suboptimal performance. In this work, we assume that the prior knowledge can be effectively represented by statistical co-occurrence correlations between entity types and relations. Based on this premise, we propose a novel algorithm called Knowledge-Embedded Graph Representation Learning (KEGRL), which enhances the representation of entity features through this statistical prior knowledge. Specifically, we calculate the statistical co-occurrence correlations existing between entity types and relations. These correlations are then ingeniously encapsulated within weighted edge-oriented heterogeneous graphs, where nodes correspond to entities and relations. Every entity node is connected to all relation nodes, and their edges symbolize the statistical correlations between entities and relations. Across these graphs, the entity features propagate under the guidance of statistical correlations, during which the statistical knowledge is injected into the entity features to enhance their distinctiveness. Extensive experiments on multiple baseline models and datasets consistently demonstrate that integrating KEGRL significantly enhances the performance of DocRE models. The code is available at https://github.com/MrDreamQ/KEGRL.

Chip-scale quantum magnetometers featuring both ultra-high sensitivity and uniform spin polarization are highly desired for practical applications and have been diligently pursued. However, the fulfillment of such capabilities for quantum magnetometers typically necessitates a separate heating unit, bulky reflector, and beyond, severely impeding on-chip integration and batch fabrication of these quantum devices. Herein, we present a novel paradigm for the wafer-level fabrication of ultra-sensitive chip-scale quantum magnetometer, which is enabled by integrating a highly reflective mirror and a temperature-controlled component on the optically transparent windows of the MEMS atomic vapor cell, thereby providing a genuinely all-in-one atomic vapor cell with a temperature stability better than ±5 mK at up to 200°C as well as a reflectivity of 95

Metal particles near insulators are the main cause of insulation failure of gas-insulated switchgear (GIS), which seriously affects the stable operation of the power grid. In this article, the experimental platform of dust movement adsorption by ac GIS was built. The effects of applied voltage level, dust material, quantity, and size on their movement behavior were studied. The special luminescence phenomenon induced by the movement of metal dust was analyzed. The results show that small size and large amount of metal dust have a greater influence on the insulation performance of the basin insulator. In the process of its movement, it is easy to form a series channel of multiple particles, which seriously distorts the electric field near the GIS basin insulator. In addition, when the charge polarity of the jumping particles is different, charge transfer will occur between the particles, and a "luminescence phenomenon" will occur. The luminescence phenomenon may directly cause the insulator to flashover along the surface, which seriously affects its insulation performance. At the same time, the "luminescence phenomenon" increases the temperature in the laboratory, creating conditions for the reaction of metal dust with oxidizing gases. This work can provide data support and a theoretical basis for the study of metal dust discharge in GIS, and the special phenomenon of "dust luminescence" found can provide new ideas for improving the reliability of GIS.

The effects of homogenization temperature and time on the microstructure of eutectic silicon in quenched cast A357 aluminum alloy, as well as the influence of solution temperature on its mechanical properties after artificial aging, were investigated. The results show that a synergistic optimization of homogenization, solution temperature and time improves the mechanical properties of the alloy greatly through optimization its microstructure with fine and spheroidized eutectic silicon and β″ precipitates with high number density. During homogenization, the eutectic silicon undergoes progressive spheroidization process with the holding time, during which the length of the eutectic silicon decreases while its width increases. It can be accelerated significantly through elevating the temperatures and will improves the material ductility greatly. Increasing the homogenization time propels the spheroidization of eutectic silicon but will also cause it to turn irregular with larger particle sizes and even transform into an amorphous state if the time is excessively prolonged due to irregular overgrowth. Elevating the solid solution temperature also produces vacancy with much higher concentration in the achieved supper-saturated solid solution after quenching, which will promote precipitation of the β″ strengthening phase during thereafter artificial aging but remains its sizes un-changed. Thereby, the precipitation strengthening potential of the alloy is utmost improved. A recommended heat treatment for the cast A357 aluminum alloy is simultaneous homogenization and solid solution at 535 °C for 12 h followed by room-temperature water quenching, and then artificial aging at 160 °C for 4–8 h.

Laser cladding (LC) technology has been widely applied in the manufacturing and remanufacturing of high-integrity surface coatings. However, the performance of the coatings is significantly affected by process parameters. To obtain coatings with optimal geometry and microhardness, this study employed response surface methodology (RSM) to explore the mechanism by which LC process parameters regulate coating performance. Single-track coatings of 316 L alloy were deposited on a 30CrNiMo8 alloy steel substrate following a design of experiments. Regression equations were established to describe the relationships between laser power, scanning speed, powder feeding speed, and the geometrical characteristics as well as microhardness of the coatings. The accuracy and reliability of the models were verified using analysis of variance (ANOVA). The results show that laser power exerts a significant influence on the formation and depth of the melt pool, with distinct melt pools forming on the substrate surface when the laser power exceeds 700 W. Moreover, process parameters affect the geometrical and mechanical properties of the coatings to varying extents. Among them, scanning speed has the most significant impact on the surface integrity of the coatings, particularly the width-to-height (W/H) ratio of the cladding track and the coating microhardness. In contrast, the width-to-depth (w/d) ratio of the melt pool and the microhardness at the interface are primarily affected by the interaction between scanning speed and powder feeding speed. This study proposes a robust experimental design and manufacturing method, demonstrating that the properties of LC coatings can be accurately predicted and controlled by adjusting key process parameters.