Accurate in situ contact temperature measurement in rocket motor exhaust plumes relies on tungsten-rhenium (W-Re) thermocouples, yet their stability is limited by oxidation-induced interfacial degradation. Here, we report an oxidation-resistant W-Re thermocouple enabled by a WSi2 interfacial transition layer and a ZrO2-Al2O3-SiO2 (ZAS) ceramic barrier coating. The resulting thermocouple achieves continuous lifetimes of 67.4, 24.1, and 6.3 min at 20 0 0, 230 0, and 270 0 K, representing similar to 25.9-, similar to 24.1-, and similar to 31.5-fold increases over an uncoated thermocouple, respectively. Across 273-2700 K, it delivers an accuracy of 1.49 %, repeatability of 1.05 %, and consistency of 0.51 %, indicating robust thermoelectric stability. In a supersonic flame test mimicking an engine exhaust plume with an average temperature of 2292.9 K, the W-Re thermocouple with WSi2/ZAS multilayer barriers operates stably for 30.5 min, confirming durability under extreme oxidizing combustion. Mechanistic studies attribute the enhanced oxidation resistance to the synergistic protection of the transition layer and composite coating. This work establishes a design strategy for stabilizing W-Re thermocouples in ultrahigh-temperature oxidative environments, enabling reliable temperature diagnostics in rocket exhaust plumes. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.
Substantial financial losses are incurred annually as a result of medical insurance fraud, driven by subtle behaviors such as false diagnoses and over-prescription. Although heterogeneous graph structures can be exploited by existing graph neural network-based methods, they often rely on manually selected meta-paths, and the identification of structural anomalies remains a persistent challenge. To address these challenges, Health insurance Anomaly Detection via Truncated Affinity Maximization (HAD-TAM) is proposed as a health insurance fraud detection method, wherein degree-normalized anomaly score-guided meta-path selection is first combined with dual-threshold edge pruning to capture structural anomalies. First, the meta-graph with the highest degree-normalized anomaly score is selected by the model to highlight structures potentially indicative of fraud. Second, non-homophily edges are iteratively pruned by the model using both global and local thresholds to suppress noise and preserve meaningful structure. Finally, the neighborhood affinity of normal nodes is maximized by the model to better separate them from anomalies and improve fraud detection accuracy. Experimental results show that HAD-TAM outperforms eleven advanced baseline methods on two real-world medical insurance datasets, achieving a 2.41% improvement in AUROC and a 1.06% improvement in AUPRC. Furthermore, it exhibits good generalization to public datasets such as Amazon and Reddit.
Persistent cloud cover and limited data availability hinder the understanding of land cover dynamics on the Southeast Xizang Autonomous Region,China.Existing studies often focus on subregions or specific land cover types,lacking long-term,region-wide assessments of change and its influencing factors.To address this gap,we integrated dense Landsat time series with the Continuous Change Detection and Classification(CCDC)algorithm,to track land cover changes from 1990 to 2020.We further analyzed vegetation trends during 2000-2020 using Normalized Difference Vegetation Index(NDVI)in combination with Sen's slope estimator.Subsequently,by integrating topographical(elevation,slope,aspect),climatic(precipitation,temperature,radiation),anthropogenic and other indicators,we employed the GeoDetector model to explore the influencing factors of land cover changes.The results showed that:(1)Land cover changed by 7.52％,with increases in cropland,forest,grassland,shrubland,and built-up land,and decreases in bareland,wetland,water bodies,and snow/ice.Bareland,cropland,and forest showed contrasting trends in the north and south;(2)NDVI increased overall(0.0008/a),with a higher growth rate in the south(0.0012/a)than in the north(0.0005/a);(3)For the three main land cover types(forest,shrubland,cropland),soil type and precipitation exhibited relatively higher explanatory power in the north,whereas in the south-with abundant moisture and strong elevation gradients-elevation and temperature played greater roles;(4)Human activities showed contrasting associations.In the north,areas with higher human activity(i.e.population density,light intensity)tended to coincide with forest expansion,and the proportion of significant NDVI increases was higher in these areas,which may largely reflect the effects of ecological restoration programs.The densely populated southern areas were more frequently associated with NDVI declines and forest degradation,driven by population growth and traditional farming practices.These findings highlight the spatial heterogeneity of land cover influencing factors and provide insights for sustainable land management and conservation in the region.
A new series of red phosphors, Ca2-xSrWO6: Eu3+(x = 0.1, 0.2, 0.3, 0.4, 0.5), was synthesized using a high-temperature solid-state technique. The phase composition, crystal structure, luminescent properties, and thermal stability of the synthesized phosphors were examined through X-ray diffraction (XRD), scanning electron microscopy (SEM), photoluminescence (PL) spectroscopy, high-temperature fluorescence spectroscopy, and fluorescence decay lifetime assessments. The findings demonstrate that Ca2-xSrWO6: xEu3+ phosphors may be efficiently stimulated by 395 nm near-ultraviolet light, resulting in pronounced red light emission at 614 nm. The light intensity predominantly arises from the 5D0 → 7F2 transition of Eu3+ at 614 nm. The luminescence intensity of Ca2-xSrWO6: xEu3+ phosphors increases initially followed by a decrease with increasing Eu3+ doping concentration. Concentration quenching occurs at a doping concentration of x = 0.3, and this phenomenon is attributed to electric dipole–electric dipole (d–d) interactions in accordance with Dexter’s theory. Moreover, the fluorescence lifetime decreases progressively with increasing Eu3+ doping concentration. The CIE color coordinates and thermal stability of the Ca1.7SrWO6: 0.3Eu3+ sample were examined. The color coordinates (0.6585, 0.3410) roughly align with the typical red light coordinates (0.670, 0.330). The color purity reached 96.5
Machine learning models for predicting mechanical properties in age-hardenable alloys often struggle with experimental datasets characterized by limited diversity with severe homogeneity, wherein aging time varies while composition and temperature remain fixed. To address this, a novel physically constrained noise augmentation strategy was proposed. Unlike standard arbitrary data augmentation, this approach introduces small perturbations strictly bounded by experimental measurement uncertainties to simulate realistic process fluctuations and explicitly break the feature invariance constraint. Using two experimental alloy datasets (age-hardenable Mg-Al-Zn and Al alloys), multiple ML regression models (Extreme Gradient Boosting, Random Forest, Support Vector Regression, Gradient Boosting Regression, and Adaptive Boosting) were trained and evaluated with and without data augmentation. Noise augmentation significantly improves predictive robustness, reducing Mean Absolute Error (MAE) by up to 30
