Na-ion batteries are promising energy storage technologies, yet cathodes suffer from structural instability during deep cycling, leading to a trade-off between energy density and long-term life. Here, we introduce a "local electron density engineering" strategy to address this intrinsic challenge. We propose that structural degradation originates from the withdrawal of electron density from lattice oxygen by a high-valence transition metal. By incorporating stable d10 (Zn2+) and d0 (Ti4+) ions, we create an electron-rich oxygen framework that acts as an "electron buffer", resisting this electron depletion. Reinforced further by Ca2+ pillars in the Na+ layers, our single-crystalline Na0.96Ca0.02Cu0.038Zn0.053Ni0.409Mn0.315Ti0.185O2 cathode exhibits a low volume change of similar to 4% under deep desodiation. In 26700 cylindrical full cells, it delivers an energy density of 181.2 Wh kg-1 and retains similar to 80% capacity rentention after 1000 cycles. These results establish a new design pathway for developing ultrastable, high-energy cathode materials for next-generation Na-ion batteries.
Assessing worker skill is essential for optimizing high-complexity, low-output production tasks that are consequently difficult to automate for increased efficiency over human work. Identifying specific, accessible metrics to identify workers as beginners versus experts may also help design more efficient training regimens, as well as help uncover the determinants of skill. In this study, atemporal (timestamp removed) eye tracking data from 16 subjects performing a soldering task were analyzed using a variety of machine learning models, including k-nearest neighbors (KNNs) and decision trees, to examine if worker task skill could be assessed from nonsequential eye movement and pupil size data alone. We further investigated whether feature extraction via principal component analysis (PCA) could be used to improve the performance, efficiency, and robustness of the prediction models. PCA was selected due to its algorithmic efficiency and previously demonstrated ability to improve the performance of tree algorithms due to the alignment of data on independent axes. We find that fine-tree models trained on 95% PCA data had the most consistent performance classifying expert and beginner sessions across unseen sessions from training subjects and sessions from unseen testing subjects (68.88% mean, 80.97% median session accuracy from training subjects; 72.64% mean, 75.13% median session accuracy from testing). This model also showed strong robustness to outlier trends, suggesting it was able to extract generalizable eye tracking trends that do not rely on sequential context but are still indicative of worker skill.
In recent years, the penetration of renewable energy sources has been rapidly increasing, and in residential photovoltaic systems, system design that incorporated self-consumption has become indispensable. To cope with this trend, single-phase three-wire inverters have attracted considerable attention. Yet, they need to support a three-terminal configuration that includes a neutral conductor. The widely used three-leg inverter topology features a simple configuration and is capable of supporting single-phase three-wire distribution systems. However, due to its two-level operation, it is associated with challenges such as noise generation and switching losses. To address these issues, we propose the NORIC topology. The proposed topology supports single-phase three-wire systems while achieving three-level operation, enabling both high efficiency and low noise simultaneously. Its effectiveness is demonstrated through simulations and validated by experimental results.
Cross-domain zero-shot anomaly detection, in which models are trained on a source domain and directly applied to unseen target domains without using any target-domain normal or abnormal samples, has gained increasing attention in image anomaly detection. Most existing zero-shot methods based on Contrastive Language-Image Pretraining estimate anomaly regions by comparing image features with predefined normal and abnormal textual prompts. Although effective in structured industrial inspection tasks, these approaches often degrade in medical imaging, where normal appearances exhibit large intra-class variability that is often difficult to represent by a single normal concept. To address this limitation, we propose a Normal-Variation-Aware cross-domain zero-shot anomaly detection framework that explicitly accounts for intra-normal variability. In addition to normal and abnormal prompts, the proposed method introduces layer-specific Normal-Variation prompts to represent ambiguous yet non-anomalous regions. By explicitly modeling normal, abnormal, and Normal-Variation states, the framework redistributes prompt probabilities more appropriately in the presence of natural normal variability, thereby reducing false positives and stabilizing anomaly decision boundaries in unseen domains. Experiments on eight medical imaging datasets demonstrate that the proposed method improves average pixel-level localization performance on datasets with pixel-level annotations, while also improving image-level detection on datasets with both normal and anomalous test samples.
