False Data Injection Attacks (FDIA) pose a serious threat to smart grid security due to their strong stealthiness. Existing methods often process spatial and temporal features in isolation or only achieve single detection functionality, making it difficult to accurately pinpoint attack sources and thus insufficient for proactive defense. To overcome these challenges, this paper proposes a novel approach that integrates Sparrow Search Algorithm (SSA) optimization with a dual-channel Convolutional Neural Network (CNN)-Transformer architecture. The main innovation of this study lies in the integration of SSA-based parameter optimization with the CNN-Transformer architecture for feature extraction, enabling unified detection and localization of FDIA: the CNN branch captures local spatiotemporal anomaly patterns induced by attacks in node measurement data, while the transformer branch explicitly models global dependencies across all nodes in both spatial and temporal dimensions via a multi-head self-attention mechanism. The SSA is introduced to adaptively optimize key hyperparameters, enhancing the model's generalization ability across power grids of different scales. Experimental results on the IEEE 14-bus and 118-bus systems demonstrate that the proposed model significantly outperforms the baseline model across multiple metrics, including Accuracy, Precision, Recall, and F1-score. Specifically, Accuracy can be improved by 0.45% to 6.86%, Precision by 0.52% to 15.21%, Recall by 1.21% to 19.12%, and F1-score by 1.13% to 17.38%.
Energy oscillation induced by the dynamic negative damping effect at the rolling interface can trigger self-excited vibration in rolling mill systems, often resulting in equipment failure and deterioration of strip surface quality. Although self-excited vibration has been extensively studied, existing models remain structurally incomplete and do not fully clarify the coupled mechanism among rolling-interface evolution, dynamic rolling-force feedback and vibration response. In this study, a novel modelling approach for rolling mill self-excited vibration is proposed by coupling the rolling-interface model, dynamic rolling-force model and mill structural model. The proposed model is used to reveal the interaction among lubrication–friction state evolution, rolling force fluctuation and vibration response. The results indicate that, as the strip thickness decreases, the contact arc shortens and the neutral-point position becomes increasingly sensitive to roll-gap and velocity fluctuations, resulting in stronger slip-zone redistribution and friction-induced negative damping. Nonlinear stability analysis further reveals a transition from a stable equilibrium to finite-amplitude self-excited oscillation. Industrial measurements show that the F3 stand exhibits the strongest horizontal vibration, with dominant frequencies of 44.922–48.828 Hz, close to the calculated horizontal natural frequency of 48.102 Hz. The predicted rolling-force deviation remains within 15%. The agreement between the numerical predictions and industrial observations supports the interpretation that the observed thin-gauge vibration is associated with the proposed closed-loop feedback among rolling-interface evolution, dynamic rolling force and structural vibration.
Energy inequality, which occurs because of the mismatch between trade-related energy consumption and economic benefits, is a key issue in global trade, challenging the sustainable development goals aimed at reducing inequalities. Whether the Belt and Road Initiative (BRI), as a major global cooperation platform, can alleviate the energy inequality hidden in global trade needs further exploration. In this paper, the embodied energy flow and value-added flow within the bilateral trading networks of 186 economies from 2009 to 2021 are calculated based on the multiregional input–output model. Subsequently, the energy inequality between bilateral trading partners is quantified using the regional environmental inequality index. Finally, implementing a staggered difference-in-differences strategy, the effect of the BRI on energy inequality and the underlying mechanisms are explored. The results are as follows: (1) Energy inequality is widespread globally and is worsening, being particularly more severe between BRI and non-BRI countries. (2) The BRI can significantly mitigate energy inequality among participating countries. Mechanism analysis reveals that policy coordination, unimpeded trade, facilities connectivity, and financial integration have notably improved since the initiative's introduction, thereby alleviating energy inequality. (3) The heterogeneity results indicate that the mitigation effect is more pronounced among the Land Silk Road countries, those country pairs with large income gaps, and country pairs with large energy endowment gaps. The findings not only offer new insights into alleviating energy inequality hidden in global trade, highlighting the significance of trade cooperation but also provide valuable references for regional cooperation worldwide.
In-situ alloying for adjusting the composition of alloys is a promising approach utilizing pre-mixed powders in laser powder bed fusion (LPBF). However, inhomogeneous composition has always hindered the application of this approach. In this study, the Ni51.8 Ti48.2 at.% alloys via in-situ alloying exposed severe cracking, due to inhomogeneous composition and inappropriate heat accumulation. To overcome these difficulties, we optimized the process parameters between volumetric energy density (VED) and layer thickness ( t ) by adjusting the scanning speed ( v ) and fabricated the crack-free, homogeneous, and dense NiTi alloys, thereby proposing a novel in-situ alloying strategy (VED -t ). Computational fluid dynamics (CFD) results demonstrate that the synergistic effects of VED and t significantly extend the melt pool lifespan and change the energy distribution of the melt pool, thus achieving compositional homogenization and porosity minimization. The strategy can also regulate the cracking tendency and crystallographic orientation by changing the melt pool type (connection, keyhole, and flat). The alloys with a B2 matrix possess a strong [001] orientation along the building direction and thus exhibit good superelasticity and elastocaloric cooling under compressive loading. These findings provide a novel approach for printing brittle alloys and drive the adoption of the in-situ alloying strategy of LPBF. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.
Cross-domain recommendation (CDR) offers an effective way to alleviate the cold-start problem, but most existing methods still model cross-domain preference transfer at the level of individual users. This point-wise paradigm relies heavily on overlapping users and struggles to capture the higher-order relational patterns underlying user interests across domains, resulting in limited performance when overlap is scarce. Meanwhile, representing items solely with ID embeddings ignores rich multimodal content, further constraining transferable preference modeling. In this work, we propose Multimodal Group-Level Relational Preference Modeling for Cross-Domain Recommendation, a framework that reformulates the transferable unit from direct user-to-user embedding mappings to user-to-key-user relational signatures, where overlapping users provide supervision and non-overlapping users serve as relational anchors. This design captures more stable and transferable cross-domain interest structures, yielding consistently improved performance. Concretely, we employ a multimodal large language model to extract item representations from textual and visual content, and further adopt a temporal attention pooling encoder to integrate users’ historical interactions with multimodal item semantics for constructing multimodal user representations. Based on these enriched representations, we develop a group-level relation learning module that models relational dependencies between each overlapping user and a domain-specific key user set selected from the non-overlapping ones, thereby enabling group-level preference transfer across domains. We further introduce an adversarial alignment strategy to reduce cross-domain representation discrepancies and enhance generalization. Extensive experiments on real-world datasets demonstrate that our method consistently outperforms state-of-the-art baselines.
