This study proposes an edge computing–based dynamic routing optimization framework to address high decision delay and poor adaptability in centralized emergency medical supply distribution during public health emergencies. Such events often cause a 300%–500% surge in medical supply demand, exposing system vulnerabilities. The framework deploys a closed-loop “sensing–prediction–optimization” mechanism at the network edge. A hierarchical analysis method quantifies the dynamic urgency of each demand point as a penalty weight in the optimization objective, while a Transformer–GRU hybrid predictor at edge nodes estimates real-time travel time and demand intensity. A proximal policy optimization (PPO) reinforcement learning algorithm enables low-latency rolling route replanning with heuristic refinement. Simulation results show an F1 score of 0.914 (95% CI [0.892, 0.936]) in on-time delivery discrimination, with an AUC of 0.967 and a top-5 NDCG of 0.934, outperforming baseline models. Compared with centralized architectures, response latency is reduced by 95.29% and weighted tardiness by 48.64%. Task completion remains above 95.5% under 50% congestion and 20% new orders, demonstrating strong robustness and the potential of edge computing and AI for resilient medical logistics systems.
With the widespread use of electric vehicles, higher standards for battery pack safety and reliability have emerged, making fault detection and diagnosis essential for stable operation. This study proposes a two-stage approach that combines a convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM) for fault detection, followed by a fault diagnosis model integrating a domain-adaptive neural network with channel attention, temporal attention, and category enhancement mechanisms. The detection model achieved maximum accuracy of 97.53%, precision of 98.03%, F1 score of 0.998, and recall of 99.31%, with minimum RMSE of 0.004 and time consumption of 49 ms, significantly outperforming comparison models. For diagnosis, the model achieved an AUC of 0.987, diagnostic accuracy of 98.33%, and time consumption of 66 ms, while demonstrating higher precision in identifying short-circuit, over-charging, over-discharging, and capacity fading faults. The proposed detection and diagnostic framework operates with high efficiency and robustness, offering reliable technical support for the safe operation and maintenance of electric vehicle battery packs.
To address the problems of inaccurate integration of multi-source information, insufficient real-time state perception, and susceptibility of dynamic scheduling decisions to local optima in port logistics operations, this paper proposes an information perception and collaborative optimization decision-making method for smart port logistics. Based on radio frequency identification technology and ultra-wideband positioning technology, the proposed method constructs a multi-source perception system integrating the identity, status, and spatial position of logistics entities. On this basis, an additive attention mechanism is introduced to enhance the ability of the long short-term memory network to capture key operational events and temporal variation features, thereby improving the accuracy of port logistics state recognition and position prediction. Furthermore, a tabu search mechanism is incorporated into the ant colony optimization algorithm to alleviate the premature convergence problem in traditional path planning and resource scheduling, enabling perception-data-driven collaborative optimization decision-making. The experimental results show that the root mean square error and mean absolute error of the proposed perception model are 1.84 m and 1.21 m, respectively, and the comprehensive classification performance index reaches 0.955. The final solution cost of the proposed collaborative optimization system is reduced to 303.28 km, with a deviation of only 1.07% from the known optimal solution. Under high-load conditions, the average vessel turnaround time is 70.32 h, while the equipment utilization rate and scheduling completion rate reach 88.90% and 85.74%, respectively. The results indicate that the proposed method can simultaneously improve information perception accuracy, path optimization quality, and resource scheduling stability in complex port operation environments, providing effective support for real-time collaborative decision-making in smart port logistics systems.
This study addresses the issues of poor selectivity, low sensitivity, and protection failure under high-resistance ground faults in existing low resistance grounding systems. We analyze the characteristics of zero-sequence current following single-phase ground faults in low resistance grounding systems and propose a high-sensitivity multi-level protection scheme based on zero-sequence current through coordinated longitudinal protection. This scheme resolves the issue of protection failure due to relatively low fault currents caused by high resistance by reducing the settings of the protection for outlet lines, branches, and boundary protections while extending their operating time. For high-resistance ground faults exceeding 1500 Ω, the scheme accurately identifies the faulted line and delays disconnection by comparing the magnitudes of zero-sequence currents at the exits of each line and the neutral line, thereby improving the accuracy of fault location. The feasibility and reliability of the proposed multi-level protection scheme for ground faults are validated through simulations conducted on a typical low resistance grounding distribution network structure.
Driven by advances in sensor technology, data-driven remaining useful life (RUL) prediction has become an important tool for bearing predictive maintenance. However, variations in operating conditions in industrial applications can cause feature distribution shifts, thereby reducing the accuracy and robustness of prediction models. To address feature distribution shifts across operating conditions, this paper proposes a graph-aware meta network based on domain generalization for bearing RUL prediction across conditions. The proposed method converts monitoring sequences into graph structured inputs and employs a temporal encoder to learn node-level degradation representations. It further integrates prior structural knowledge, globally shared structure, and task-adaptive structure to build multi-level dynamic adjacency relationships, and introduces an attention pooling mechanism to obtain graph-level degradation representations. During training, a meta-learning strategy with inner and outer loop updates is designed to dynamically adapt the graph generation parameters, enabling task-specific topology to adjust automatically with changing operating conditions and thereby improving model generalization. Cross-condition experiments on a public bearing accelerated life dataset demonstrate that the proposed method achieves superior prediction error metrics across multiple transfer tasks and delivers stable performance under varying operating conditions.
Metrology has become the inevitable guarantee to gain confidence in all areas of industry and science. In the specific field of ballistics, significant technological progress has been achieved in recent years in terms of measurement means and techniques, including high-speed data acquisition, improved piezoelectric pressure sensors, advanced digital signal processing techniques and modern optical velocity measurement systems. These developments contribute to improving the reliability and performance of both weapons and ammunition, thereby mitigating potential risks to human life and economic losses. However, the traceability of measurement results is not always obvious due to the absence of primary standards for certain physical quantities or the difficulty of estimating measurement uncertainties. Within the framework of the accreditation of the ballistics laboratory according to the requirements of the ISO/IEC 17025 standard, significant efforts have been undertaken to contribute to the establishment of a well-founded approach for the assessment of uncertainties of the main measurable quantities in ammunition proof tests, namely the ballistic pressure and the flight velocity of the projectile using piezoelectric transducers and light screens. In addition to the technical aspects of measurement, this work aims to highlight the factors that significantly affect the uncertainty of projectile velocity and gas pressure measurements. First, possible sources of measurement errors are determined, then their elementary contributions to the uncertainty are estimated. Finally, global uncertainties are evaluated based on the propagation of variances as mentioned in the guide to the expression of uncertainty in measurement (GUM). Furthermore, the validity of these uncertainties has been verified using the Monte Carlo method in accordance with Supplement 1 of the GUM. The required simulations were performed by the LNE-MCM software (Laboratoire National de Métrologie et d’Essais – Monte Carlo Method). The results obtained allow us to implement a practical method for estimating the uncertainty of measurement of the most sought-after ballistic quantities, considering their importance in evaluating the performance of firearms and their ammunition according to standards in the field of ballistics.
With the gradual opening of low-altitude airspace and the rapid development of UAV technology, large-scale UAV swarms are increasingly used in logistics, inspection, and security scenarios. To achieve accurate and robust tracking of highly dynamic, high-density, and strongly interactive UAV swarms, this study proposes a radar-based interactive multi-model multi-target tracking algorithm. Point cloud quality is improved by integrating velocity vector density clustering with adaptive constant false alarm rate detection. An interactive multi-model framework incorporating uniform velocity, uniform acceleration, and coordinated turning modes is established, together with a group potential field–driven state transition mechanism. Adaptive thresholding and dynamic track splitting and merging strategies are further introduced to enhance tracking stability. Experimental results show that the proposed method achieves an average distance RMSE of 1.47 m and a speed RMSE of 1.18 m/s, representing reductions of 49.8% and 42.3% compared with the traditional joint probability data association algorithm. The average tracking accuracy reaches 88.19% and remains 84.31% under a clutter density of 80 points/scan, while the average computation time is 36.92 ms, satisfying real-time requirements. The results demonstrate improved accuracy and stability for low-altitude UAV surveillance in high-density urban scenarios.
This paper studies the finite-time synchronization control problem of multi-manipulator systems under aperiodically intermittent communication. In contrast to the traditional periodic intermittent communication schemes widely adopted in previous research, the introduction of aperiodic patterns breaks through the limitations of regular communication intervals. This novel approach more realistically mimics the complex and unpredictable communication conditions often encountered in actual industrial and robotic applications. By employing the appropriate auxiliary variables and establishing a velocity estimator, a coordinated tracking controller is designed to realize the finite-time synchronization. Note that the settling time monotonically increases with the maximum rest ratio that the system can tolerate. Finally, the validity of proposed finite-time synchronization strategies is verified through a numerical simulation.
There are high-risk problems, such as peripheral rock instability and palm face collapse, during tunnel construction, and the traditional monitoring methods are difficult to meet the safety management needs due to sparse data and lagging response. This paper proposes an intelligent monitoring system for the whole process of tunnel construction based on Internet of Things (IoT) and digital twin, which integrates a multi-source sensor network, BIM dynamic modeling, and risk intelligent analysis. The system realizes an all-around perception of environment, equipment, and surrounding rock status through real-time fusion of heterogeneous data, and uses digital twin technology for 3D visualization and risk trend prediction. It adopts the improved D-S evidence theory for multi-source risk assessment, and improves the early warning accuracy through the effectiveness factor and conflict weakening strategy. The actual engineering experiments show that the system achieves 100% monitoring coverage and 100% warning accuracy, and successfully captures the whole process of palm face collapse and the time-sequence evolution of enclosing rock deformation, which significantly improves the safety and management efficiency of tunnel construction. The study verifies the high accuracy and stability of the proposed system, which provides an intelligent solution for the safety of complex underground projects.
With wide application of new materials, structures, and technologies in modern industry, testing objects were no longer limited to conventional materials and common shapes. Demands of testing new materials and complex shapes bring challenges to researchers in non-destructive testing area. Combining robots, which have been widely used in industry, with the non-destructive testing technology can replace manual operation and improve testing precision. Additionally, the robot-assisted systems can enhance efficiency and safety of the testing process. The researchers have carried out numerous designs and implementations to combine robots with non-destructive testing devices. This article presents four non-destructive testing systems with robot, including single-arm robot holding a transducer, single-arm robot holding a tested part, and twin-arm robot holding two transducers. In conclusion, application scopes are analyzed to help users select appropriate systems according to sizes, materials and defect types.
To address the impact of reflection boundaries and port and starboard ambiguity on passive sonar positioning, a method of vector hydrophone hybrid source positioning under different reflection boundaries is proposed. When the reflection boundary is close to or located within the near-field Fresnel zone, the reflected waves will increase and decrease because of the superposition of the line-of-sight waves, thus affecting the positioning results. The composite vector hydrophone consists of three-dimensional orthogonal vector channels and scalar channels. By fully utilizing the vector information in the MUSIC algorithm array flow pattern, the ambiguity problem of the azimuth angles on both sides under far-field and near-field reflection boundary conditions can be solved. By establishing different reflection models, the array manifold matrix of the three-dimensional directional angle and distance of the sound source under dual reflection is derived to illustrate the influence of reflection boundaries on position estimation. Comparing the simulation results with Cramer Rao bound (CRB) and maximum likelihood estimation algorithms, it is proven that the proposed method is correct and efficient. Finally, the effectiveness of the method is verified through lake experiments. This study has important guiding significance for the practical promotion of underwater acoustic engineering.
Robotic actuators need to be light weight, compact, and efficient for meeting the requirement of size and controllability. It's capability of achieving high power density is often restricted by limitations in torque output and efficiency. This paper presents a novel multi-module spliced direct-drive outer rotor BLDC motor tailored for robotic systems. This study focuses on a unique multi-module splicing structure that simplifies manufacturing and assembly while significantly enhancing torque density through improved magnetic symmetry and inherent structural modularity. Critical electromagnetic parameters—pole-arc coefficient, air gap, and permanent magnet thickness—are systematically refined using combined theoretical modeling and high-fidelity simulations in ANSYS Maxwell. A quasi-Newton multi-objective optimization algorithm accelerates convergence toward globally optimal configurations, effectively balancing multiple design objectives. Optimized results confirm a peak efficiency of 95.06%, core and copper losses reduced to 22.5 W and 12.8 W respectively, a 38.89% reduction in cogging torque, an 8.33% decrease in air-gap flux density, and torque ripple maintained at 24% in simulations, primarily attributed to the 12th harmonic. Prototype testing validate these improvements, with actual torque ripple slightly higher than the simulated value at 26%, while demonstrating agreement with simulation data in efficiency and losses. By integrating structural design, computational optimization, and experimental verification, this work delivers a robust solution to the persistent high torque density versus manufacturing feasibility trade-off, enabling more efficient, reliable, and scalable robotic actuations.
With the increasing demand for higher accuracy and reliability in the field of engineering measurement, traditional methods have shown a series of problems when facing complex scenarios and precision measurement tasks. Therefore, a scale invariant feature transformation engineering measurement method integrating binocular vision is proposed. This study focuses on binocular vision three-dimensional dimension measurement, using two-dimensional chessboard for monocular and binocular calibration to obtain internal and external reference information. At the same time, the scale invariant feature transformation algorithm has been simplified, combined with epipolar geometry to improve matching performance. The experiment showed that the improved scale invariant feature transformation algorithm achieved a matching accuracy of 97%. After fusing binocular vision, the close range matching was improved to 98%, the matching time was reduced to 1.8 seconds, and the number of feature points was reduced to 24. In distance measurement, the minimum error for planar targets was 0.25%, the maximum error for curved targets was 1.08%, and the overall maximum error percentage was 2.24%. The scale invariant feature transformation operator that integrates binocular vision has achieved significant results in engineering measurement, showing higher accuracy and reliability in three-dimensional dimension measurement compared to traditional methods. This innovative method is expected to improve measurement accuracy and reliability, providing a more accurate and feasible solution for three-dimensional dimension measurement in the engineering field.
This paper aims to verify the issue of accuracy and consistency in measuring marine humidity by introducing a comparison of humidity measurement carried out by three research institutions. The focus is on aspects such as measurement value traceability, humidity sensor calibration method, comparison method, and analysis of the comparison results. Based on uncertainty analysis, the study further examines the impact of distinct calculation models for reference values. Moreover, the results of the humidity comparison demonstrate a high level of consistency among the three marine metrology laboratories, providing support for the accuracy as well as reliability of the national marine humidity measurement data. Additionally, this humidity comparison work provides valuable insights for research on reference value.
This paper focuses on the MKE1620A CNC cylindrical grinder, using an equivalent analysis of the joint spring-damping characteristics to investigate the overall performance of the grinder and propose directions for design optimization. A total of 168 spring-damper elements were established to model the fixed, movable, and bearing joints. These elements were classified and calculated to determine their parameters, which were then incorporated into a finite element analysis model to examine the impact of joint stiffness on the static and dynamic performance of the machine tool, with results showing less than a 10% error compared to actual measured data. Additionally, the paper investigates how the quality of six common structural materials influences the first-order natural frequency of components and explores the relationship between variations in joint stiffness and changes in the machine tool’s natural frequency. The findings provide theoretical and data-driven insights for the design and optimization of CNC cylindrical grinding machines, serving as a valuable reference for enhancing machine tool performance and machining quality.
With the development of intelligent manufacturing systems, data-driven fault diagnosis has become a hot research topic. Traditional data-driven fault diagnosis methods often rely on expert-extracted features, wherein feature extraction process requires considerable effort and affects the final results to a great extent. However, end-to-end fault diagnosis methods based on deep learning can automatically learn feature representations from raw data. In this study, first, the raw vibration signals of various fault states of a planetary gearbox were segmented and preprocessed into input table data types. Second, a convolutional neural network with wide kernels in the first layer was used to extract gear fault features from the raw vibration data of the gearbox. Then, the multi-head attention mechanism was incorporated to focus on different feature spaces and obtain diverse feature information. Finally, using the Softmax layer, the fault features were classified and fault diagnosis of the gearbox was achieved. Validation experiments and comparative analysis indicated that the proposed fault diagnosis model exhibits stronger learning ability as well as a simpler and convenient diagnostic process compared with the traditional methods. The proposed model has broad application prospects in data-driven fault diagnosis.
Microfluidics is a rapidly growing technology with applications in biochemistry and life sciences. To support the ongoing growth there is a need for common metrology, quality control, and standardisation. Here measurements of wettability and surface roughness can contribute, and these quantities affect flow characteristics of devices, bonding processes in manufacturing, and special microfluidic mechanisms such as droplet formation and spreading of fluids on surfaces. To quantify wettability, an optical laboratory setup was used to measure liquid drop contact angles of three liquids on a microfluidic surface. To further quantify wettability, the Owens, Wendt, Rabel, and Kaelble model was applied to contact angle measurements to determine the total surface free energy. To quantify surface roughness, atomic force microscopy and stylus profilometry measured area roughness parameter and profile roughness parameter for four samples of microfluidic surfaces. The wettability methods successfully demonstrated measurements of contact angles, and these methods were applied to determine a value for the total surface free energy. AFM and stylus profilometry successfully determined surface roughness parameters, and the determined values agreed with the expected for the material. In conclusion, the demonstrated methods can contribute to metrology, quality control and standardisation in microfluidics.
This paper presents an IoT (Internet of Things) based smart building fire cloud monitoring system to enhance fire safety in smart buildings. It integrates low-cost sensors and real-time video surveillance for real-time environmental data collection. Data are uploaded to the cloud for remote monitoring via a custom web interface. The system features an artificial neural network model that reduces computational complexity and response time, achieving >95% accuracy in fire prediction. It assists in planning evacuation routes based on fire location, enhancing safety and efficiency. Laboratory and field tests confirm reliable performance, and the novel system will find applications in smart fire detection and prevention.
An improved grey model (GM) for predicting free-form surface machining errors was established to address the low measuring efficiencies of coordinate measuring machines (CMMs) and the low prediction accuracy of the GM(1,1) model. A number of points on a free-form surface are measured with a CMM, and machining errors are obtained. Ideas from metabolic methods and the GM(1, N ) prediction model were combined. To reduce the impact of random fluctuations of the machining errors, a Markov prediction model was then used to correct the fitted results for the residuals and obtain the predicted results of the machining errors for free-form surfaces, thus improving the prediction accuracy of the model. The predicted results of the metabolic GM(1,1) and metabolic GM(1, N ) models were then compared. The experimental results showed that the combination of metabolic theory, a grey model, and Markov theory effectively improved the prediction accuracy.
In the context of the European Tempus project “Quality in Higher Agricultural Education in the Mediterranean (QESAMED)”, Analysis and Characterization Center of Cadi Ayyad University has committed to implement a quality approach according to the ISO/IEC 17025 standard. The first objective was to accredit the testing carried out in the microbiology laboratory in response to strong demand from the food industry in the Marrakesh region. The process began with an initial assessment of the center activities to identify the main gaps from ISO/IEC 17025. The range of laboratory and standard testing to be accredited were determined and an appropriate action plan was established. After that, a staff training was programmed to improve their skills in relation to this standard and metrology concepts. Implementation of the metrology function is one of the key steps for the deployment of the continuous improvement process. This function guarantees the traceability of measurements and the reliability of microbiological testing results. Several actions have been carried out, including: (i) identification of critical quantities and associated metrological requirements, (ii) checking of the metrological consistency of equipment through the calculation of the capability coefficient. The management of equipment requires several steps, from receipt of the equipment to its decommissioning or reform: (i) identification of the equipment, (ii) creation of equipment files, (iii) performance of calibration, verification, and maintenance operations. Ultimately, the definition of a strategy ensuring metrological traceability will optimize management costs by taking into account the structural constraints linked to the organization of metrology on a national scale.
