Due to the presence of the singular matrix in discrete-time singular systems, state variables cannot be directly updated, which makes the traditional optimal control method difficult to apply directly. This paper investigates the finite-horizon linear quadratic optimal control problem for discrete-time singular systems with stochastic multiplicative noise and multiple input delays. To address this issue, the paper constructs an augmented matrix to transform the discrete singular stochastic system with multiple time delays into a standard discrete stochastic system while ensuring its regularity and causality. Then, based on the maximum principle, the optimal control conditions are derived, and the optimal control law and minimum cost function are obtained using coupled Riccati difference equations. Finally, simulation results demonstrate that under the influence of random disturbances and input delays, the proposed method effectively stabilizes the system state, verifying its effectiveness.

The open-world scene graph generation task has difficulty in detecting unknown objects and their relationships. To address this issue, a relation-reasoning model based on multidimensional feature fusion (MDFF) is proposed. The proposed model is combined with an open-world object detector to form a two-stage open-world scene graph generation algorithm. First, the pretrained open-world object detector identifies objects in the input images. The MDFF model then performs relationship inference based on the detection results. Comparative experiments are conducted on the VG 150 dataset using traditional methods and the MDFF model. The experimental results indicate that the MDFF model achieves 7% improvement in recall rate for predicate classification tasks. Moreover, the performance improves by 3% in open-world scene graph generation and zero-shot inference tasks. Furthermore, ablation studies have confirmed the effectiveness of different feature dimensions on model performance improvement.

To address the issue that traditional path planning algorithms cannot effectively solve the multi-objective problem of Automated Guided Vehicle (AGV) in path planning, the standard whale optimization algorithm is improved. Tent chaotic mapping and adaptive nonlinear dynamic inertia weight are introduced into the standard whale optimization algorithm, and the convergence factor and search coefficient are improved. Then, the improved algorithm is combined with the A * algorithm for multi-objective point path planning. The iterative curve and running time of the improved whale optimization algorithm are tested using standard test functions, and a simulation comparison is conducted between the improved whale optimization algorithm and the standard whale optimization algorithm in the same map environment. The results show that, with a fixed population size, the improved whale optimization algorithm has a faster convergence speed and search accuracy compared to the standard whale optimization algorithm.

In cloth-changing person re-identification tasks, clothing variation is a critical factor degrading recognition accuracy. To discover clothing-invariant features, a Positional Mask-Guided Model (PMGM) for Cloth-changing Person Re-identification was proposed. The PMGM model leverages four positional masks (head, upper body, lower body, and arms) to guide the network in capturing local fine-grained features, which are fused with global features to precisely extract clothing-invariant representations. During inference, integrating head feature matching with identity feature matching further enhances the model discriminative capability. Experimental results show that the PMGM model achieves 5.7% improvement in mAP and 6.1% improvement in Rank-1 on the PRCC dataset compared to baseline models.

To solve the problem of class-based grasping in multicategory tiled scenes, this paper adopts different feature fusion methods and proposes a joint optimization algorithm MC-GSNet (Multi-Class GraspNet) that fuses category semantics and grasping posture and an optimization algorithm MT-GSNet (Multi-Task GraspNet) that builds a multitask learning model. The improved methods explicitly incorporate category information, optimize the generation logic of grasp poses and enhance the algorithm’s adaptability and success rate in multi-category object planar scenes. Experimental results on the public dataset GraspNet-1Billion demonstrate that the proposed methods significantly improve task adaptability and grasping success rates in multi-category planar scenes. MC-GSNet and MT-GSNet achieve 32.6% and 43.9% average accuracy improvements in grasp detection, respectively; MT-GSNet exhibits superior adaptability to unseen objects due to its integration of segmentation features. The experimental results in the simulation environment show that the grasp successful rates (GSR) of MC-GSNet and MT-GSNet reached 88.3% and 95.0% respectively, which can meet the needs of actual engineering deployment.

To address the low energy harvesting efficiency of conventional energy harvesters under low-frequency broadband vibration excitation, an internal resonance piezoelectric energy harvester with time-varying potential wells (TI-PEH) was proposed. TI-PEH consists of a piezoelectric cantilever beam, an inverted auxiliary beam, a magnetic pendulum, limiters, a magnet at the free end of the inverted auxiliary beam, and a mass block. The elastically supported magnet is introduced to create time-varying potential wells, enhancing the coupling effects between internal resonance and nonlinear magnetic interactions. The experimental results show that within the range of 2~5 Hz, compared with the traditional structure, the maximum root mean square (RMS) voltage of the TI-PEH under 0.1 g excitation intensity is increased by 276%. Under random excitation conditions, the coupling effect between the time-varying potential well and the internal resonance can significantly enhance the output performance of the TI-PEH. Furthermore, stiffness reduction of the auxiliary beam enables the TI-PEH's operational bandwidth to shift toward lower frequencies, demonstrating superior energy harvesting performance in ultralow-frequency vibration environments.

To address the challenges of increased production space, equipment transportation and installation costs caused by the excessive volume and mass of forging hydraulic press centering devices, this study focuses on the lightweight design of an 80MN forging hydraulic press centering device. A 3D model of the centering device was developed using SolidWorks, and static analysis was performed with ANSYS Workbench finite element software based on its actual stress conditions. Four design variables were selected for optimization using the response surface methodology. A mathematical model was established with minimum total mass as the objective function, constrained by maximum deformation and equivalent stress. Static reanalysis of candidate points demonstrated that optimized results met deformation and stress requirements while achieving a 1 284 kg mass reduction. The lightweight design effectively reduces device weight, improves material utilization, and enhances overall production efficiency while maintaining required strength and stiffness.

To address the issues of fragmented knowledge and limited information sharing in the field of workpiece machining distortion, this paper proposes a knowledge graph framework tailored for this domain. A dataset comprising 1 085 long-text records related to machining distortion was constructed, and an ontology model was developed to define the relevant entities and relationships. A BERT+BiLSTM+CRF-based entity extraction model and a BERT-based relation extraction model were employed to enable automated knowledge extraction. Knowledge fusion was performed through a combination of cosine similarity-based matching and manual verification. A knowledge graph containing 4 330 entities and 5 509 relationships was then built using the Neo4j graph database, with support for data insertion, deletion, updating, and querying implemented via Cypher language. Experimental results demonstrate that the proposed knowledge graph can effectively support decision-making and research analysis in workpiece machining processes.

To address the issues of a prolonged design cycle, high costs, and limited optimization effects associated with traditional giant tire side panels design that rely on empirical formulas and experimental validation, the side plate of a 33-inch large tire vulcanizing machine as the study object was used, and finite element analysis (FEA) methods and software such as ANSYS were employed. A three-dimensional model was constructed using UG, and a casting cavity design was used to replace the conventional solid structure. Various thicknesses and arrangements were configured, and meshing and material parameter settings, followed by mechanical and thermal analyses. The results indicate that when the distance between the casting cavity and the inner surface of the side plate is set to 80 mm, approximately 14.72% of material is saved. Additionally, the heating performance of six evenly distributed casting cavities is optimal, with 48.05% increase in heating speed compared to designs without casting cavities. This approach not only satisfies the structural performance requirements of the side plate but also reduces manufacturing costs, offering effective theoretical and practical guidance for the lightweight design of large tire side plates.

To optimize traffic management and reduce environmental pollution through scientific methods, a method for constructing vehicle driving conditions based on an improved K-means++ clustering algorithm is proposed. Combined with Markov chain theory, this method analyzes and constructs vehicle driving conditions. The collected vehicle driving data are preprocessed, including data cleaning and feature extraction. Dimensionality reduction was performed using Principal Component Analysis (PCA), and a K-means++ algorithm based on cosine similarity is introduced. The optimal number of clusters is determined using the elbow method. The results show that four driving conditions effectively simulate real driving scenarios. The comparison of average silhouette coefficients from the clustering results demonstrates that the improved algorithm significantly outperforms traditional methods in clustering performance. Using the Markov chain model, the transition relationships between the driving condition states are validated, and the final vehicle driving conditions are constructed. According to the comparative results of the relative error of key characteristic parameters, the average relative error is only 4.726%, indicating that this method has high rationality and accuracy in simulating actual road conditions. This provides a solution for traffic data analysis and model construction in complex traffic environments.

In order to maintain the handling stability of electric vehicles under different road conditions and ensure the good power performance and steering ability of vehicles. The fuzzy control theory was used to identify the maximum adhesion coefficient of the current road surface and determine the optimal slip rate of the road surface. Aiming at the optimal slip rate, the output torque of the motor is adjusted based on the sliding mode control theory to realize real time tracking of the wheel slip rate and maximize the power of the vehicle. The acceleration slip regulation control of dual-motor four-wheel drive vehicle was designed by Simulink, the vehicle model was built by Carsim, and the feasibility of the control was verified by joint simulation. The results show that the error of vehicle identifying the optimal road slip chamber is less than 7% when the drive acceleration slip regulation is involved. When the left and right road surface of the vehicle changes, the maximum lateral displacement and maximum steering wheel angle of the driving acceleration slip regulation control vehicle are 40% and 15% of the non-control vehicle, and the lateral stability optimization is obvious.

A thorough understanding of the pollutant emission characteristics of heavy-duty diesel vehicles is of great significance for the upgrading of the city's mobile pollution source monitoring methods. This study obtains the actual road emission data and driving condition data of heavy-duty diesel vehicles in Qingdao through remote monitoring methods. By using big data analysis techniques, the impact of specific power on the NOx and CO2 emissions of diesel vehicles was analyzed, as well as the temporal distribution characteristics of pollutant emissions. The results show that the contribution rates of NOx and CO2 pollutants under acceleration, constant speed, deceleration, and idle conditions decrease in sequence. Under uniform acceleration conditions, the emission factors of NOx and CO2 are relatively high in the low-speed condition range, slightly decrease in the medium-speed condition range, and continue to decrease in the high-speed condition range. Appropriately increasing the average driving speed of heavy-duty diesel vehicles can reduce their NOx and CO2 emissions. The pollutant emissions of heavy-duty diesel vehicles show a trend of being higher during the day and lower at night, with the peak emission period of NOx and CO2 being from 5:00 to 8:00 every day.

In view of the current problems such as insufficient accuracy of ionospheric real-time and forecast data and the difficulty in fusing multi-source heterogeneous data, this paper designs a set of ionospheric fusion processing and forecasting system based on Global Navigation Satellite System (GNSS) and vertical observation data, aiming to provide high-precision and high-timeliness ionospheric environment information services for radio system applications. This system adopts Limited Kalman Filter Model, based on the Kylin operating system and cloud computing platform. It utilizes container cloud, high availability, and distributed architecture to achieve high-precision real-time and forecast reports of the total electron content (TEC), critical frequency of the F2 layer (foF2), and electron density in the global and surrounding areas of China. Experimental tests show that the system has a current reporting delay of approximately 5 minutes and a spatial-temporal resolution of 5°×2.5°×15 min, representing a significant improvement over traditional data processing methods. Additionally, the system supports three-dimensional electron density visualization, providing reliable data support for applications such as ionospheric research, satellite navigation correction, shortwave communication, and ground-based radar. It offers high-precision and high-timeliness ionospheric environment information services for radio system applications.

In response to the late start and weak research foundation of alpine skiing in China, based on motion data from inertial sensors, a scientific motion analysis and training guidance method for alpine skiing is studied. By analyzing key characteristics such as the athlete’s center of gravity trajectory, upper limb stability, and blade angle, specific technical points are guided, and the cosine distance between training movements and standard movement data is calculated to complete the similarity evaluation of movements. A data analysis system for alpine skiing movements is developed to complete real-time movement data transmission and display, as well as historical movement data analysis and evaluation, and to vividly display the training effects of athletes through a visual interface. This study provides data support for optimizing training strategies and improving athlete performance for coaching teams.

Surface defect detection is a key aspect of quality inspection of industrial components. Aiming at the lack of systematic literature research on surface defect detection algorithms for industrial parts, China National Knowledge Infrastructure (CNKI) and WOS (Web of Science) core ensemble databases are selected as data sources between 2017 and 2023. With the help of CiteSpace visual analysis software, the research line of surface defect detection algorithms in the field of industrial components inspection is analysed by the number of annual publications and keyword clustering. The current state of research on deep learning-based algorithms for detecting surface defects on industrial components is systematically presented, as well as the practical applications of single-stage and two-stage target detection algorithms. The key problems of current surface defect detection algorithms for industrial components and the corresponding solution strategies are summarized. The future development of surface defect detection algorithms for industrial components is also discussed.