Real-time pulse monitoring could enable the diagnosis of cardiovascular conditions and thus has potential application value in cardiovascular health assessment and long-term physiological status monitoring. However, current pulse sensors are predominantly limited to point-pixel signal acquisition, which fails to emulate the 3D spatial perception inherent in traditional Chinese medicine, posing a significant challenge for subsequent diagnosis. Herein, a wearable laser-induced graphene (LIG) pulse sensor array was developed. Before fabrication of LIG, precision mechanical processing was utilized to mill an array of groove structures on the polyimide (PI) substrate to achieve stress-concentration structures, thereby enhancing sensitivity. Then, the obtained porous LIG framework was modified with MoS2 through a dip-coating process. Unexpectedly, the MoS2 nanosheets filled the micropores or cracks of porous LIG and mitigated cross-unit variations within the sensing array, achieving excellent batch-to-batch consistency across different sensor arrays. Furthermore, a novel chemical bond between PI and outer polydimethylsiloxane (PDMS) was developed for flexible encapsulation to prevent the adverse effects of atmospheric humidity on the functional properties of LIG and thus ensure long-term stability. The as-fabricated sensors exhibited several superior properties that are characterized by enhanced sensitivity (2.205 kPa−1) in the low-pressure range (0–15 kPa), excellent consistency (maximum interdevice error of 6.61
Fault diagnosis for hydraulic manipulators plays a crucial role in ensuring operational safety but still faces challenges in fault localization. To address this issue, a guided differential dilated convolutional network (GDDCN) is proposed in this study. First, a novel interference masking mechanism is designed to provide dual guidance for feature extraction and classification. Then, a learnable differential kernel with the center parameter fixed at zero and side parameters opposite in sign is designed to adaptively extract gradient features. Afterward, a multiscale gated dilated convolution (MGDC) module is developed to capture global temporal features across multiple scales and achieve gated feature fusion. Finally, the fused features are fed into a fully connected classification module for fault classification. The results show that the GDDCN achieves diagnosis of the faulty joint with an average accuracy of 99.39
Integrated energy system (IES) scheduling optimization involves multi-energy coupling, multidevice coordination, and complex constraints. Conventional approaches rely on experts to manually formulate optimization models, resulting in high technical barriers, lengthy development cycles, and limited transferability. To address requirement uncertainty, attention drift in long contexts, and high debugging costs when applying large language models (LLMs) to IES scheduling optimization, this study proposes the MASEO, a multi-agent system for energy optimization. MASEO decomposes the optimization workflow into four sequential stages—information collection, mathematical modeling, code implementation, and execution verification—each handled by a dedicated LLM agent. A structured checklist is introduced to standardize information collection, while an error-classification-based traceable repair mechanism routes errors to the responsible agent for targeted correction. Case studies on a Beijing data center IES and a German community IES show that MASEO achieves results highly consistent with expert benchmarks, with an annualized cost deviation of approximately 1.4
Gastrointestinal (GI) drug delivery remains challenging because sustained constant-rate infusion and positional stability are difficult to achieve simultaneously. Conventional magnetic capsules often rely on burst-type or nonlinear release mechanisms. To address these limitations, a magnetically controlled dual-target constant-rate drug delivery capsule system (M-DCDC) is presented. The system enables highly linear and quantitatively adjustable drug infusion through specialized screw–piston transmission. This mechanism establishes a direct relationship between the magnetic rotation frequency and the delivery rate, achieving a maximum flow rate of 0.1 mL/s with predictable dosage control. Sequential dual-target delivery is achieved through functional decoupling. By reversing the direction of the magnetic field, two independent reservoirs, each with a volume of 0.6 mL, can be selectively actuated at separate target sites. To resist intestinal peristalsis, a self-locking mechanical anchoring module is integrated to maintain positional stability without continuous power consumption. Analytical modeling, finite element analysis (FEA), benchtop tests, and ex vivo experiments using porcine intestines demonstrate that M-DCDC can withstand a peristaltic force of 0.32 N while achieving localized staining. By combining stable anchoring with decoupled constant-rate infusion, the proposed system offers an effective engineering approach for targeted treatment of multifocal GI diseases.
针对火电厂煤炭输送与卸载场景中的粉尘、 噪声、 遮挡和工况变化导致的安全隐患识别延迟及可追溯性不足问题, 构建秒级响应、 规则可解释的多模态安全监测系统。 1. 提出知识图谱增强的多模态监测框架, 融合视频、 传感器和文本日志, 实现规则约束和可审计告警; 2. 设计由视觉异常、 传感器阈值和知识图谱规则共同驱动的混合触发机制, 降低边缘视觉漏检导致的风险; 3. 构建 2018 2024 年 235000 个对齐样本的数据集, 并通过低秩自适应、 安全感知近端策略优化、 半监督学习和长尾蒸馏实现工程部署。 1. 从规程、 手册、 事故日志和专家知识中构建安全知识图谱, 并将实体/关系嵌入注入多模态模型; 2. 利用触发中心滑动时间窗对视频和传感器进行时空对齐, 并采用置信度感知融合生成风险判断; 3. 将 72B 教师模型蒸馏为 8 位量化 7B 边缘模型, 并在边云协同架构中完成现场部署。 1. 系统在两周现场试验中产生 3652 条告警, 严重隐患专家验证精度为 90
The large-scale deployment of high-energy-density lithium-ion batteries (LIBs) in electric transportation and grid storage has imposed increasingly stringent requirements on battery safety, reliability, and intelligent management. However, the limited observability of internal electrochemical, thermal, and mechanical states remains a fundamental challenge, leading to persistent safety risks, degraded low-temperature performance, and accelerated aging, which collectively hinder the scalable adoption of electrified systems. To overcome these challenges, conventional battery management systems (BMSs) are evolving beyond voltage-current-temperature measurements toward high-fidelity state estimation enabled by advanced nondestructive sensing technologies, including emerging internal and implantable diagnostic concepts. Based on a comprehensive analysis of physical signals associated with material aging and failure mechanisms, this review provides a systematic and critical assessment of nondestructive testing techniques for battery state monitoring. Beyond a conventional technique-oriented summary, recent advances in battery state diagnosis and lifetime management algorithms are examined, with a particular emphasis on multisource physical feature fusion strategies. More importantly, this review establishes a unified framework linking degradation mechanisms, internal physical signals, and state estimation strategies, offering a cross-scale perspective to guide the codesign of advanced sensing technologies and intelligent algorithms, thereby facilitating the development of next-generation BMSs.
In recent years, flexible magnetic sensors have undergone rapid development in cutting-edge fields such as wearable devices, robotic hand interaction, and soft robotics, demonstrating remarkable performance in sensitivity, multi-dimensional force perception, and deformability. The core component of these sensors is a flexible magnetic film, fabricated by blending magnetic particles with a soft polymer matrix using specific forming processes. Flexible magnetic force sensors work by decoding the link between magnetic signals and mechanical forces to sense touch. This review systematically examines the research framework of flexible magnetic sensors for force sensing from four critical dimensions. First, the manufacturing processes are analyzed, including the preparation of soft magnetic functional films and commonly used magnetosensitive elements. Subsequently, two primary force-decoupling methodologies are discussed: analytical modeling and data-driven approaches, with a comparative evaluation of their respective strengths and limitations. The review summarizes the key performance characteristics of flexible magnetic sensors, focusing on sensitivity, multi-dimensional force detection and deformability. Furthermore, emerging applications in wearable devices and robotic hand interaction are highlighted, illustrating the capability of these sensors for high precision feedback and multi-axis force decoupling. Finally, we assess several key challenges. These include performance variation due to non-uniform magnetic particle distribution, which necessitates laborious calibration, and signal drift following repeated use. Future directions, including bioinspired designs and artificial intelligence enhanced system integration, are also outlined. By synthesizing these perspectives, we aim to advance research toward higher sensing performance and broader application horizons for flexible magnetic sensors.
Thermal batteries based on phase change materials (PCMs) are a key technology for energy savings and carbon reduction in building heating since latent thermal energy storage by PCMs with high energy storage density indicates great potential to utilize renewable energies. However, the complicated structure of finned-tube heat exchangers and the nonlinear melting–solidification process of PCMs require a great deal of computation and time resources, highly restricting the engineering design and application of PCM-based thermal batteries. To address these issues, this study proposed a universal machine learning-enabled performance prediction framework for finned-tube PCM-based thermal batteries designed for residential domestic hot-water supply, in which thermal energy is stored in PCM during the charging process and released to cold water during the discharging process to provide usable hot water for end users. First, a simplified simulation method coupling a 1D tube model with a 3D computational fluid dynamics model is established for the rapid performance computation of thermal batteries. Second, the deep operator network framework is introduced to directly map static parameters to the time-based temperature response of outlet hot water validated by the simulation and previous experimental results. Based on the machine learning-enabled performance prediction framework, the effects of critical parameters such as tube outer diameter and thickness combination, fin distance, and flow rate on the heat storage capacity and heat release power are systematically analyzed, providing guidelines for the proposed flow rate feedback regulation strategy of thermal batteries to satisfy different usage temperatures Tuse and time constraints. The results show that the proposed framework can accurately predict the outlet temperature and available total volume of hot-water output by thermal batteries under various conditions.
The evolution of next-generation district heating networks toward higher efficiency and sustainability is constrained by persistent challenges in supply–demand coordination and lifecycle optimization. This paper provides a comprehensive review of artificial intelligence (AI) integration across four key lifecycle stages of heating networks: planning and design, construction and renewal, operation and control, as well as maintenance and fault diagnosis. Critical research gaps are identified, including application fragmentation and persistent data silos. The literature has demonstrated that AI-based approaches can deliver significant performance improvements across multiple lifecycle stages. In the planning phase, AI can improve design computational efficiency, in some cases by up to an order of magnitude. In the construction phase, AI-enhanced management has been shown to accelerate project timelines while reducing costs. For operational control, deep learning models can reduce thermal load forecasting errors by more than half. In the maintenance phase, AI enables multiday early fault warnings with localization accuracies exceeding 95
Bioinspired robotic fish can provide compliant, low-noise underwater locomotion, but compact shape memory alloy (SMA)-driven designs often suffer from slow response and poor fatigue life. We propose a grass-carp-inspired pectoral–caudal-fin (PCF) soft robot to realize body caudal-fin (BCF) and median paired-fin (MPF) multimodal swimming and specifically improve SMA performance via embedded thermal management. The robot integrates an SMA spring caudal-fin actuator and a large-strain SMA pectoral-fin actuator within a watertight compliant body and adopts a sealed-cavity liquid-cooling scheme. Experiments show markedly enhanced actuation bandwidth and durability, enabling stable multimodal locomotion with a 45.67 mm/s peak speed and 9.12 (°)/s peak turning rate, plus ultrasonic-based obstacle avoidance at 20 cm. The results suggest a compact pathway to resilient, versatile underwater soft robots for long-duration operation in cluttered environments.
碳中和目标正在推动现代能源系统由集中、 单一和确定的运行模式向分布式、 多能耦合和高度不确定的形态加速演进。 风能、 光伏、 储能、 氢能及供热网络的大规模发展对能源系统的预测、 规划、 调度、 控制和全生命周期管理提出了更高要求。 人工智能可通过数据驱动预测、 物理信息学习、 强化学习、 智能优化、 数字孪生和生成式模型加快高保真模拟, 提升可再生能源预测与消纳能力, 促进电-热-气-氢协同调度, 并服务于设备设计、 故障诊断、 材料发现、 碳核算和排放监测。 本专辑聚焦人工智能与低碳能源工程的深度融合, 强调将物理机理、 不确定性量化、 安全约束和可解释决策嵌入智能方法, 推动能源研究由黑箱预测迈向可信决策、 由单一设备优化迈向系统与全生命周期协同。 未来需进一步解决数据共享、 模型泛化、 网络安全、 隐私保护及工程验证等问题, 为构建安全、 经济、 韧性和可验证低碳的未来能源系统提供科学依据与技术支撑。
Although high-speed train (HST) bogie covers effectively reduce aerodynamic drag, they raise safety concerns due to lift-induced oscillations caused by unsteady underbody flow. This study investigates the effects of various bogie covering structures on aerodynamic load pulsations, pressure fluctuations, and underlying flow mechanisms in HSTs using an improved delayed detached-eddy simulation at 400 km/h. Three covering configurations are considered: fully enclosed covers (FECs), separated-type covers (STCs), and skirts-only. The numerical results show that all covering configurations significantly reduce the total aerodynamic drag. Specifically, FECs achieve the largest reduction at 19.85
Modular reconfigurable underwater robots (MRURs) have attracted increasing attention due to their potential to overcome the limitations of conventional underwater robots with fixed structures and single-task capabilities. This review summarizes recent progress in MRUR from three perspectives: modular structural design, sensing and perception, and control strategies. For structural design, this review examines module division, housing, and connectors, highlighting composite pressure-resistant structures and magnetic docking mechanisms as feasible solutions. For perception, MRURs require not only multi-sensor fusion but also topology awareness, docking perception, and distributed perception to handle changes in reconfiguration. For control, MRUR reconfiguration alters topology, dynamics, and actuation; therefore, key issues include topology recognition, docking and separation control, propulsion redundancy, distributed coordination, and the transition toward self-reconfiguration. Finally, representative application scenarios, key technical challenges, and future research directions of MRURs are discussed. With continued technological progress, MRURs are expected to evolve toward self-reconfiguration and self-adaptation, achieving intelligent control and cooperative autonomy.
Electric drive systems (EDS) feature high integration and strong nonlinearity, which renders weak fault identification a challenging task. Acoustic particle velocity signals employed for noncontact fault monitoring are prone to contamination by noise and redundant features. This contamination seriously interferes with weak fault extraction and reduces diagnostic stability. To overcome the limitations of conventional single or multiple feature selection strategies, this paper proposes an information entropy-based multisource feature fusion selection (IE-MSFS) method. The proposed method can effectively eliminate redundant information and enhance the characterization ability of weak fault features. Based on the EDS acoustic particle velocity signals collected in the laboratory, a comparative analysis with vibration signals is carried out on three typical weak fault types through machine-learning evaluation. The results verify that the proposed fault diagnosis method scheme exhibits outstanding and stable weak fault recognition performance, with a diagnostic accuracy exceeding 95.2
This paper proposed a numerical modeling method for generating ultrahigh porosity aluminum foam through independent control of pore size and cell wall thickness. The proposed method introduced two independently adjustable parameters (point spacing and scaling distance) to control pore size and cell wall thickness, respectively. Aluminum foam models with a porosity exceeding 90
Efficient fuel-air mixing within milliseconds is critical for scramjet performance, yet the flow physics of a transverse jet in supersonic crossflow remains insufficiently quantified. Planar Rayleigh/Mie scattering and stereoscopic particle-image velocimetry were applied to a Mach 2.68 crossflow. Jets with a 2 mm orifice diameter were injected at three dynamic-pressure ratios (1.64, 4.92, and 8.19) under two incoming boundary layer thicknesses (1 and 4 mm). Instantaneous imaging captured the bow shock, barrel shock, Mach disk, slip line, recirculation zone, and counterrotating vortex pair (CVP). Vorticity fields revealed streamwise vortices forming beside the barrel shock, merging 20 mm downstream from the jet orifice, and persisting as a typical CVP that entrained freestream fluid. Boundary layer thickness systematically enhanced jet penetration and modulated near-field breakup patterns without altering far-field mixing limits. The penetration depth was fitted using a modified correlation, indicating an approximately 10
Remolded moraine soils are highly susceptible to degradation caused by freeze–thaw (F-T) cycles, posing significant risks to the stability of slopes in hilly regions. This study investigates the mechanical degradation and microstructural changes induced by F-T cycles through a comprehensive experimental program, including F-T cycle tests, triaxial shear tests, unconfined compression tests, and micro-computed tomography (micro-CT) scanning of moraine soil from the southeast of the Xizang Autonomous Region, China. The results reveal three stages of mechanical parameter degradation: rapid decline, moderate reduction, and plateau. Cohesion, initial average elastic modulus, and strength decrease significantly under low confining pressure (40
Tunnel overbreak is a common but unfavorable phenomenon in drill-and-blast excavation, leading to increased construction costs, delayed schedules, and potential stability risks. Accurate evaluation and prediction of overbreak are therefore important for tunnel construction control. In this paper, we propose a cloud-model-based comprehensive evaluation and prediction framework developed by integrating fuzzy evaluation theory with subjective and objective weighting methods. A dataset containing 523 records from the HuXiTai (HXT) Tunnel was used, and seven routinely obtainable geological and blasting indicators were selected to construct the evaluation system. Eight weighting strategies were compared, including conventional objective methods, expert judgment, and a ridge-regression-based objective method. The best-performing objective weights were further combined with subjective weights to establish the final comprehensive evaluation model. The results show that the proposed model achieved an overbreak evaluation accuracy of 85.28
The dynamic behavior of memristive Josephson neurons is highly sensitive to circuit configurations. This study proposes a novel control method for neuronal circuits based on targeted parallel shunting using an external capacitor. A hybrid circuit integrating a memristor, a Josephson junction, and a nonlinear resistor is constructed, with an external capacitive branch introduced in parallel with the nonlinear resistor to achieve precise manipulation of neuronal firing patterns. It is demonstrated that the external capacitive branch enables effective regulation of the firing patterns of the neuronal circuit through its targeted parallel connection to the nonlinear resistor. By continuously adjusting the external capacitance parameter, the system can realize controllable switching among various firing modes. Concurrently, variations in the stimulus amplitude reshape the internal energy distribution framework of the system, determining the dominant roles of different energy storage components. These two mechanisms constitute a dual-dimensional “mode–energy” regulation system for the neuronal circuit. Furthermore, the regulatory mechanism of the external branch originates from its unique local shunting effect and specific energy exchange process. The energy evolution of the external capacitor exhibits dynamic characteristics distinct from those of the main system, and this asynchronous energy response can effectively perturb the global balance of the system. The proposed method provides a foundation for the precise control of neuronal dynamics in memristive Josephson systems.
Subaqueous dunes hold significant research value in sedimentary geomorphology and marine engineering. Based on two phases of multibeam bathymetric surveys conducted in the western offshore region of Hainan Island, China, this study systematically examines the geometric characteristics and dynamic evolutionary mechanisms of subaqueous dunes. The results show that dune formation is governed by multiple interacting factors, including hydrodynamics, sediment transport processes, and water depth, with tidal currents serving as the primary driving force for dune initiation and growth. Dune evolution is jointly modulated by the interplay between bedload transport and suspended sediment concentration. The lag effect of suspended sediments represents a key mechanism contributing to dune development, while dune morphological stability is constrained by suspended sediments as well as by limitations imposed by water depth. Specifically, crest erosion suppresses further increases in dune height, whereas water depth regulates the critical scale of dune growth. Subaqueous crescentic dune height plays a dominant role in controlling morphological stability and hydrodynamic response, whereas dune width exhibits strong linear relationships with both height and wavelength, highlighting its integrative role in linking vertical and horizontal morphological development. The growth of crescentic dunes proceeds through four successive stages: initiation, chasing, merging, and splitting. Throughout this evolutionary sequence, sediment supply plays a crucial role in regulating dune wavelength and shaping the overall morphology. Correspondingly, dune geomorphology evolves sequentially from crescentic to curved, bifurcated, and ultimately linear forms. This study offers theoretical support and practical insights for understanding seabed geomorphic evolution, regional sediment dynamics, and marine engineering applications.
