This PDF file contains the front matter associated with SPIE Proceedings Volume 13395, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

A weighted vertical difference contrast measure (WVDCM) for small infrared maritime target detection is proposed for the complex background of the sea surface and the difficulty of infrared weak target detection under environmental interference conditions. The algorithm firstly adopts a multi-scale vertical mean difference filtering method with weighted information entropy to pre-process the images. Secondly, the multi-scale patch local contrast method with human visual system (HVS) is utilized for enhancing the detection target while suppressing the background, and ultimately, the adaptive threshold segmentation method is employed to acquire the actual target. The results of the simulation experiments of this algorithm on the sea surface data-set show that the image’s get higher detection rate under the condition of low false alarm rate, and the signal-to-clutter ratio gain (SCRg) and background suppression factor (BSF) of the image are more obviously improved than other algorithms.

Fibrous porous material has complex special network structure and porosity feature. Effective digital description of these complex micro structures is a challenge that needs to be solved in engineering fields. Building digital model for active design is the basis of function design and manufacturing optimization of related materials. In this paper, a new digital modeling method that aims at metal fibrous porous material is explored based on the growth model viewpoint and using Monte Carlo method. It means to build a parametric feature model which can map the forming processes of these complex micro structures. This feature model makes advances in geometric authenticity and parameters controllability by introducing porosity, gravity and interference as feature control parameters. By comparing the feature attributes with the reverse scanning reconstruction model and combined with fluid simulation experiments, the efficiency of active design model for describing metal fibrous porous structures is verified.

The variational method is used to investigate the parameter evolution characteristics of incoherent dual beams in cubicquintic competing nonlinear media. It is found that the coupling between the two soliton beams directly affect the soliton’s transmission speed and phase. The cubic and quintic competing nonlinearities are observed to have no impact on the transmission speed and amplitude of the two soliton beams, meaning that the amplitude is maintained constant during propagation. Additionally, the cubic and quintic competing nonlinearities do not influence the shift in the center position of the two soliton beams. The rate of phase change with respect to the transmission distance is shown to be linearly related to the cubic and quintic nonlinear coefficients of the medium. Both cubic and quintic competing nonlinearities are found to increase the rate of phase change with transmission distance, with quintic nonlinearity having a significantly greater effect on the phase change rate compared to cubic nonlinearity. This difference becomes more pronounced as the soliton amplitude is increased.

3D printing technology has brought significant changes to traditional manufacturing industries and is crucial for the development of precision machinery, computers, CNC, lasers, new materials, and other fields. However, it still cannot completely replace traditional manufacturing industries. With the application of laser 3D printing technology in nontraditional industrial fields, this technology not only breaks through the limitations of traditional processing and manufacturing, but also improves the performance of traditional 3D printing technology and realizes characteristic manufacturing, and has been successfully applied in rapid manufacturing systems. This article takes various classifications of laser 3D printing technology as an example to illustrate its current status and application methods in the manufacturing field, and looks forward to its future development prospects in the field of rapid prototyping manufacturing.

The aim of hyperspectral image (HSI) classification is to define categories for the different labels that are assigned to each pixel vector. Generative adversarial network (GAN) can mitigate the limited training sample dilemma to some extent, but there are still two critical issues, namely balance collapse and insufficient sample diversity. In this article, we propose a Coupled Dual-Channel Generative Adversarial Network for HSI classification. It mainly consists of a Coupled Generative Network (CGN) and a Dual-Channel Discriminative Network (DDN). CGN achieves the reconstruction of HSI samples through cascaded convolutional layers, in which the label information of the samples is used to avoid the balance collapse, while DDN extracts spatial attention weights and spectral attention weights of the input true/false samples respectively, and performs feature mining in both spatial and spectral dimensions in a dual-channel style. In order to reinforce the detailed features of input samples in both spatial and spectral dimensions, we design a new Cascaded Spatial-Spectral Attention Block (CSSAB). Finally, feature maps at different scales are fused for final sample discrimination and classification, which can mitigate the effects of insufficient sample diversity. Experimental results on two HSI data sets demonstrate that the proposed CDGAN effectively improves the classification performance compared to some state-ofthe-art GAN-based methods

Accurate attitude estimation is critical for the navigation systems of ballistic missiles. The SINS/CNS is commonly utilized in missiles. When noise parameters are unknown or changing over time, a single Kalman filter cannot meet the accuracy requirements. Hence, this paper explores an improved multi-model adaptive estimation (IMMAE) algorithm incorporating weight parameters to improve the accuracy. The optimized estimation is obtained through a weighted sum of sub-filters with different weight parameters. Additionally, the IMMAE algorithm effectively monitors noise and exhibits rapid adaptability of weight parameters under varying conditions. The simulation results demonstrate superior performance and enhanced adaptability of IMMAE algorithm compared to traditional Kalman filter and MMAE algorithms.

Augmented Reality (AR) enhances user interaction with digital content through real-world overlays, finding applications across various fields. AR glasses, serving as an excellent AR platform, have been developed in both optical see-through and video see-through types. Professional video see-through devices and advanced optical see-through devices with vision systems can perform environment recognition and hand detection but are often bulky and heavy for prolonged wear. Conversely, lightweight optical see-through AR glasses, which lack embedded systems and have limited sensors, serve primarily as displays. While they offer the advantage of reduced weight, they lack advanced interaction capabilities. In this research, we utilize an Android mobile phone as the computing unit and present an interactable framework for AR glasses with limited sensors. This framework supports head motion estimation, hand gesture detection and tracking, providing a robust AR experience without the need for high-end hardware. It has been tested on lightweight optical see-through AR glasses only equipped with an Inertial Measurement Unit (IMU) and single camera. Our solution offers a cost-effective and portable approach, enhancing data visualization and virtual object operation.

Due to the complexity of temporal and spatial characteristics during microwave heating, the unpredictability of the magnetic field makes it difficult to accurately predict the heating effect by traditional modeling methods. Aiming at the problems of uneven material temperature distribution and unpredictable heating results during microwave heating, a variable frequency strategy is proposed to improve the heating uniformity and estimate the heating effect using a support vector machine. Firstly, this paper proposes variable frequency heating to move the hot spot position according to the fixed frequency heating effect. Then the particle swarm optimization (PSO) algorithm support vector machine (SVM) model is introduced to predict the uniformity index of variable microwave heating. The numerical calculation results show that the proposed method has accurate prediction results for the heating results of the variable strategy.

To address the issues of low resolution and blurry edges in infrared images of substation scenes, a super-resolution reconstruction method based on a dual-attention guidance mechanism is proposed. Specifically, during the deep feature extraction of infrared images, the spatial and channel transformer group (SCTG) is proposed to extract global spatial similarity features, fully utilizing the long-range dependencies of non-local pixels and enhancing detailed information. Subsequently, a Spatial Frequency Information Fusion Module (SFIFM) utilizes the extracted high-frequency information, reducing artifacts and mosaic effects that occur during the super-resolution process. The overall quality of the reconstructed images is improved and the detailed contour information is refined. Finally, ablation and comparative experiments on a self-made dataset demonstrate that the proposed method outperforms state-of-the-art methods.

In this study, we thoroughly investigated the impact of the calendering process on the microstructure of lithium-ion battery electrodes and explored the related fracture behavior and heat transfer characteristics. Through numerical simulations and experimental validation, we systematically analyzed the effects of different calendering parameters on the contact force and heat transfer efficiency between particles within the electrode. The results indicate that by appropriately adjusting the calendering parameters, we can optimize the electrode microstructure, increase the particle compactness, and thereby enhance the electrode's stability and mechanical integrity. These in-depth insights provide a substantial theoretical foundation for the manufacturing and design of lithium-ion batteries, potentially offering strong guidance for the development and improvement of future battery technologies.

With the scaling down of feature size, the proportion of region sensitive to single-event effect (SEE) with respect to active region increases. The simulation challenges the anti-radiation technology for space application. Silicon-on insulator (SOI) technology has been utilized for radiation hardened integrated circuits. This work takes advantages of TCAD tool to simulate SEE in SOI NMOSFET, focusing on the effects of linear energy transfer (LET) of injected heavy-ion, top silicon film thickness, drain bias, and floating body effect on single event transients (SET) pulse. Mechanisms are investigated, which provides guide for radiation hard SOI technology.

In order to address the structural limitations inherent in the current grey prediction model, this paper proposes the time power conformable fractional derivative non-linear grey Bernoulli model (CFNGBM(p,1) model) based on the traditional grey Bernoulli model, leveraging the theory of fractional calculus. Firstly, by utilizing the computational advantages of conformable fractional derivatives, the grey derivative of the traditional whitening equation is extended from first-order to fractional-order, making the model structure more flexible. Secondly, a time power term is introduced into the model structure to fully capture the non-linear relationship presented in real systems. Additionally, to address the overfitting phenomenon in grey models, an extrapolation optimization mechanism is incorporated into the traditional parameter optimization process by simulating future prediction scenarios. Finally, a case study of monthly production data of photovoltaic glass in the crystalline silicon industry chain demonstrates that the proposed CFNGBM(p,1) model outperforms three other grey Bernoulli models in terms of accuracy, and the use of extrapolation optimization mechanism significantly reduces overfitting.

This paper discusses the sensor technology to improve the indoor positioning ability of educational assisted robot. Under the background of mechanics, by integrating various sensors (such as inertial measurement units, optical sensors, etc.) and using particle filtering algorithm, more accurate indoor positioning is realized. The study adopts modular design, which divides the robot system into three subsystems: robot, action recording and wireless remote control, and ensures performance through independent development and testing. In the positioning system, combined with odometer, IMU and lidar data, the dynamic particle number particle filtering algorithm is used to achieve global accurate positioning. This study is of great significance for improving the application effect and personalized educational experience of educational-assisted robots.

There is an obvious angle between the outer and inner wings of large birds during flight, gliding and landing, and the corresponding angle is different under different flight conditions. In this paper, the influence of different outer and inner wing angles on aerodynamic characteristics is studied. Based on the existing bat-like flapping-wing aerial vehicle (FAV), a series of wing models with different outer and inner wing angles are established, then the aerodynamic characteristics are analyzed by XFlow-Abaqus co-simulation, and the simulation results are verified by wind tunnel experiment. The results show that lift increases first and then decreases with the increase of the angle between the outer and inner wing, and the maximum lift can be obtained when the angle is 15°, compared with the angle of 0°, the lift increase is about 10%, the thrust and the pitch moment decrease with the increase of the angle. The research in this paper provides theoretical guidance for the wing structure optimization of flapping wing robots, and is helpful to the wing structure design and optimization of flapping wing robots.

Robot system is the key technology for the development of space on-orbit service technology. In view of the relative movement between the manipulator and the target during the capture process, the force sensor is used to obtain the interaction force and torque information between the gripper and the docking ring, and the manipulator opens the zero force control to adjust the pose. At the same time, according to the contact force information between the gripper and the docking ring obtained by the tactile sensor, whether the dual-arm gripper is closed is judged. In this paper, the system modeling and zero force acquisition control of space manipulator are studied under the research background of space failed satellite acquisition mission. Based on the ground microgravity air flotation simulation system, the experimental results show that the space dual-arm zero force capture control system based on force/contact fusion is feasible and it has excellent performance.

In the current society, the application of fire-fighting explosion-proof elevators is reflected in more and more scenarios, indicating that fire-fighting explosion-proof elevators have great development potential. In this paper, the rough analytic hierarchy process (RAHP) and the mass function development method (QFD) are used to study and analyze the design of large-load and low-energy fire-fighting explosion-proof elevators, and put forward design suggestions. Firstly, the user demand index was obtained through market research and other methods, and the weight value of each demand was calculated by using the RAHP method, and then the user demand was transformed into technical indicators and the quality house was established, and finally the quality house was analyzed and conclusions. The RAHP-QFD method can well combine the user's needs with the technical indicators, and can more intuitively reflect the relationship between the user's needs and the technical indicators. At the same time, the combination of the two methods can complement the shortcomings of the two methods as a strong basis for conclusions.

Revealing the dynamic characteristics of the temporal and spatial evolution of false alarms is of great significance for decision support to reduce the high false alarm rate of the transmission line monitoring systems. Therefore, based on the alarm data extracted from the transmission line monitoring system, various methods such as coefficient of variation, Kernel density analysis, and spatial correlation were utilized in this study. The evolution characteristics of the spatiotemporal pattern of alarm at the city and district levels were analyzed. The results show that: (1) Through the coefficient of variation (CV), we found that the coefficient of variation (CV) of alarm showed an upward trend at the city levels. (2) Through the kernel density analysis, the median of the false alarm increased. The largest rate of false alarms is minhang, baoshan, and fengxian. The maximum rate of the false alarm changed from baoshan to minhang. (3) Through the rank-scale rule, it is found that the numbers of districts in HH and LL regions are larger than the other two regions, and they remain unchanged. The districts in HH are mainly concentrated in the western region and the districts in LL are mainly concentrated in the middle region.

Due to the adaptive transmission parts hydraulic torque converter, compact planetary gear speed mechanism and proportional valve precise control of clutch control system, AT transmission has the advantages of good road adaptation, outstanding mobility characteristics of starting and accelerating, high power density, and superior shift comfort. Therefore, it is widely used in military, mining and other working vehicles with complex road conditions and requiring high mobility flexibility. The matching calculation of AT transmission and vehicle parameters plays a vital role in the dynamic performance of the vehicle. This paper conducts the dynamic calculation and real vehicle verification of a vehicle carrying AT to ensure that the dynamic performance of the vehicle meets the design requirements.

Autonomous decision-making has the advantages of strong information processing ability, fast response, and high costeffectiveness, which is a key technical support for achieving unmanned, clustered, and intelligent combat. Firstly, clarify and sort out the basic concepts of autonomous decision-making in unmanned clusters and the development history of major countries and regions. Subsequently, in line with the trend of modern combat style transformation, combined with the main technical methods, the basic working principles of autonomous decision-making were analyzed in depth, mainly involving multi-source information fusion, battlefield situation assessment, task allocation optimization, and cluster path planning. Finally, the future development trends of autonomous decision-making methods were analyzed from the perspectives of enhancing intelligence, improving adaptability, enhancing system stability and sound security.

Addressing the inherent uncertainties in the linearization of diesel engine nonlinear models and the susceptibility of conventional PID speed regulators to external perturbations, this study endeavors to enhance the control strategy through the application of robust control theory. A systematic analysis of the robust stability of the uncertain dynamical system leads to the formulation of a H∞ controller design, which is then refined through the solution of Linear Matrix Inequalities (LMI). Simulation experiments are conducted employing MATLAB/Simulink integrated with a detailed nonlinear diesel engine model. The outcomes demonstrate that the proposed robust control approach significantly outperforms traditional PID controllers in terms of its ability to accommodate modeling discrepancies and mitigate the effects of external disturbances, culminating in an improved precision in speed regulation.

In order to solve the current problems of low material availability rate for coal machinery equipment overhaul, which leads to long equipment overhaul period and high overhaul cost, through analyzing the reasons and problems caused by low material availability rate, it is proposed to optimize supplier delivery control , accurate production planning, detailed material classification management and setting up department KPIs and other measures to improve the material completeness rate. At the same time, we introduce constraint-based material re-planning and ABC material classification methods, and focus on key materials and key materials to comprehensively Improve the company's material management system and lay the foundation for high-quality and efficient development of the company.

Bionic rotary tillage universal blade is a kind of universal blade. it could complete both broken stubble and rotary tillage. In this experiment, the single-cutter torque (power consumption) of bionic rotary tillage broken stubble was tested by orthogonal test method. The main factors affecting power consumption were found out by using the method of multiple variance analysis. The rotational speed and depth of the universal blade for bionic rotary tillage were determined. The operation speed of the unit has no significant effect on the power consumption of the universal bionic rotary tillage broken stubble blade. The mathematical models of power consumption, blade speed and tillage depth are established by using regression analysis method. The power consumption of rotary tillage and broken stubble operation could be predicted by using the established mathematical model.

To address the limitations of conventional vibration analysis in detecting faults in low-speed heavy-duty rolling bearings, this paper employs acoustic emission technology to capture fault-related information during their operation. Subsequently, the gathered acoustic emission signals are decomposed using the CEEMDAN (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) algorithm. By incorporating the correlation coefficient and variance contribution rate, a sensitive IMF (Intrinsic Mode Function) component is selected, and its energy entropy is calculated as the extracted fault feature. This feature is then fed into a neural network for accurate fault classification. To validate the effectiveness of this approach, a simulation test platform for low-speed heavy-duty bearing detection was established. Experimental results demonstrate that this method achieves high accuracy and promising outcomes in addressing the fault identification challenges of low-speed heavy-duty rolling bearings.

The isolation switch in Gas insulated Switchgear (hereinafter referred to as GIS) serves as the critical power equipment within the power grid system; however, an effective method for detecting the mechanical state of the isolation switch is hitherto lacking. Hence, this paper investigates the detection approach for the mechanical state of the GIS isolation switch. This paper focuses on the research of a model 252 kV GIS isolation switch, simulating typical mechanical faults such as rust sticking, closing not in place, and three-phase inconsistency. The study utilizes a wireless automatic torque wrench detection system to collect torque-angle data during the opening and closing process under various simulated mechanical conditions. Key feature information such as mean torque of stationary section, peak torque of rigid minute/junction point, and Angle of overrun torque in the torque-angle curve are extracted for analysis of the characteristic laws of operating torque and rotation angle under different mechanical states. Finally, a mechanical state detection method for GIS isolation switches based on characteristic parameters of torque and angle is proposed to provide reasonable suggestions for the operation and maintenance of isolation switchgear.

This paper proposes a zero-voltage switching (ZVS) control strategy for a four-switch Buck-Boost (FSBB) converter that is easy to calculate and minimizes the peak value of the inductor current. Different from the complex calculation formula used in related studies to calculate the Root-Mean-Square (RMS) value of the inductor current, the proposed control method can equivalently achieve the control goal of minimizing the RMS value of the inductor current by calculating a simple expression for the peak value of the inductor current. At the same time, since no complex calculation is required, the computing power requirements for the microcontroller can be reduced, and the controller can complete real-time calculations at a higher frequency to achieve a more realistic and accurate control effect. In addition, the control strategy proposed in this paper only needs to store a small number of boundary limit curves. Since there is no need to store key calculation results, the requirements for the microcontroller FLASH space can be reduced. A 400 W FSBB converter simulation is performed to verify this method.

This article proposes a design scheme for a marine garbage cleaning robot based on the Internet of Things (IoT) and Attitude Heading Reference System (AHRS) to address the increasingly serious problem of marine garbage pollution. The robot achieves precise positioning, posture measurement, and intelligent recognition by integrating multi-sensor information, optimizing the garbage cleaning process. The research adopts STM32 as the main control module, combined with three-axis gyroscope, three-axis accelerometer, and three-axis magnetometer, and provides accurate attitude and heading information through AHRS algorithm to assist garbage cleaning operations. In addition, the robot is equipped with image acquisition equipment, underwater cleaning devices, etc., which can perform garbage recognition, grabbing, and cleaning in nearshore areas. The user interface is intuitive and easy to use, supporting real-time monitoring and data feedback, improving work efficiency and safety, and providing strong support for marine environmental protection.

In municipal pipeline engineering, the retaining of ultra-deep structures often employs a cylindrical structure composed of an underground wall. However, existing engineering manuals and literature primarily discuss the retention of polygonal excavations, with relatively insufficient research on cylindrical support structures, which leads to issues in the vertical reinforcement configuration in practical engineering. This paper aims to address this issue by establishing a computational model for the active soil pressure on a polygonal approximate circular continuous wall, considering the self-stabilizing effect of the force arch. The experimental results indicate that the distribution of soil pressure in circular shafts is influenced by factors such as the depth-to-diameter ratio, soil layer properties, and groundwater seepage. Moreover, the derived analytical solutions allow for a more accurate calculation of soil pressure coefficients, thereby optimizing the design of underground continuous walls for shafts.

To enhance the precision of greenhouse air temperature regulation, this paper introduces an optimization approach for fuzzy Proportional-Integral-Derivative Controller (PID) control parameters utilizing the Grey Wolf Algorithm (GWO). Initially, a dynamic mathematical model for temperature control is formulated. Subsequently, GWA is employed to finetune the three key parameters of the fuzzy PID controller: proportional gain (Kp), gral gain (Ki) and derivative gain (Kd) thereby identifying the optimal controller settings. Further, the MATLAB/Simulink platform is leveraged to conduct comparative simulation studies against traditional PID control, fuzzy PID control, fractional-order PID control methodologies. The paper culminates with an evaluation of the improved GWA against other optimization techniques such as the Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Bat Algorithm (BA) for the parameter tuning of the fuzzy PID controller. The findings indicate that the fuzzy PID control system, refined through the enhanced GWO, exhibits swift response characteristics, minimal overshoot, commendable robustness, and robust stability.

This paper investigates the problem of fixed-time tracking control for a space manipulator subject to internal uncertainties and unknown disturbances. To shorten the system’s response time, a novel globally fast fixed-time stable system is first developed. Based on this system, a novel non-singular terminal sliding mode surface is designed, which ensures fast and fixed-time convergence regardless of the initial states. A robust fast fixed-time sliding mode controller is then constructed by combining an adaptive mechanism, which can guarantee the tracking errors converge quickly to small regions around the origin within a bounded time. With the proposed control method, there is no required to know prior information about the bound of the lumped uncertainty. The suggested scheme is analysed using the Lyapunov stability theory, and the effectiveness is demonstrated through numerical simulations.

In order to study the influence of welding on the aluminum alloy box girder, the welding temperature field of the aluminum alloy box girder was simulated by using the finite element software Abaqus, and then the stress change of each plate of the aluminum alloy box girder during the welding process was analyzed. The results show that the bearing capacity of the aluminum alloy box girder is reduced by 24.94% by welding, and the bearing capacity of the steel box girder is reduced by 16.67% under the same conditions, indicating that the impact of welding on the aluminum alloy box girder is more obvious. When the aluminum alloy box girder reaches the ultimate bearing capacity, the failure mode is the same as that of the steel box girder, that is, bending failure.

Three-dimensional (3D) reconstruction of real-world scenes is a crucial task in various fields such as virtual reality, computer graphics, and urban planning. With the advancement of technology, the combination of Neural Radiance Fields (NeRF) and Unmanned Aerial Vehicles (UAVs) has gained significant attention for efficient and accurate 3D reconstruction. This paper presents a comprehensive discussion on the technical pathway for integrating NeRF with UAVs to achieve real-scene 3D reconstruction. The proposed approach leverages the capabilities of deep learning, computer vision, and aerial robotics to produce detailed 3D models of real-world environments. Mathematical formulations and algorithms are presented to demonstrate the feasibility and effectiveness of the NeRF-UAV integration in 3D reconstruction.

The emerging application of robots in household scenarios provides a promising solution for daily housekeeping, warehousing and healthcare. In such scenarios, the robot needs capabilities for environment perception, object searching, navigation, and object manipulation. However, existing solutions often focus on a single functionality and cannot handle complex and unseen environments, imposing a huge barrier to collaboratively processing information in complex environments. In this paper, we propose a holistic robot control framework that allows a robot to percept, search, navigate, and grasp target objects at specified locations, providing a complete solution with the major functionalities for smart home robots. To tackle the issue of complex background interference in the environment, we developed a perception module that combines instance segmentation with point cloud clustering to extract spatial information of the target object. Based on the spatial information obtained by the perception module, we constructed an object-searching algorithm that uses related objects as clues that accelerate the object-searching process in unseen environments. In addition, a new grasp detection model is presented that explicitly considers the information of both the objects and backgrounds to reduce the interference of background on grasp detection, improving the grasp success ratio. Results from simulation experiments demonstrate that the proposed approach significantly outperforms the baseline method in terms of both efficiency and success rate.

In the instant of start-up, idle or stop, sounds that with high frequency and particularly harsh, will be heard by the passengers inside the car. According to the mechanism of noise, it can be divided into three categories: electromagnetic noise, mechanical noise and aerodynamic noise. This paper creates a noise test system based on LabVIEW, for more in-depth analysis of the noise. Through the analysis of the test data, error of this system is less than 1%, in line with the general design requirements. In addition, it has a superior function than the traditional sound level meter. Specific functions are as follows: (a) cheaper; (b) display of noise waveform; (c) data preservation; (d) capture amplitudes beyond the rated point; (e) frequency spectrum analysis

In recent years, with the construction and development of computer technology, sensing technology, identification technology and 5G network in our society, Microcontroller Unit (MCU) technology has entered a rapid development stage, constantly changing the production and lifestyle of our society. On this basis, production and lifestyle are constantly developing towards comfort and intelligence, which has become a social trend. Based on this background, an access control system with single-chip microcomputer as the core of the system is designed. The core component of this access control system is STM32f103vet6 single chip microcomputer, which uses human infrared, touch screen, steering gear, RFID and other sensors and modules to realize the functions of password, IC card identification and fingerprint identification. And alarm function of buzzer and LED lamp; The opening and closing function of the door lock. On this basis, a voice broadcast function is developed, which can automatically broadcast the reserved information of visitors. At the same time, WeChat applet is developed for remote access.

Composite blades, known for their exceptional specific strength, stiffness, and damping characteristics, are pivotal in addressing blade fatigue issues in aero-engines. This paper designs and manufactures eight composite fan blades with various lay sequences, examining the effects of ply ratio and lay sequence on the blades' vibration characteristics. A vibration test platform was constructed, subjecting the eight blades to varying levels of sinusoidal, random, and impact excitations. The vibration responses were measured using an accelerometer (A3), a displacement sensor (X1), and a strain gauge (S1). The measured blade vibration modes and natural frequencies were compared with finite element simulation results. The study reveals that blades without 90° plies exhibit significantly higher natural frequencies than those with 90° plies. The comparison shows that the error between simulation and experiment is less than 8% for the first five modes, demonstrating the finite element simulation model's effectiveness. Additionally, different stacking sequences at the same ply ratio influence the natural frequencies. The first vibration mode frequency decreases with increasing excitation magnitude, indicating a certain degree of nonlinearity in the blades.

This paper explores constellation-level autonomous mission planning techniques for distributed network, aiming to address the complexity of mission planning brought about by the increase in the number of remote sensing constellation satellites in the rapidly developing commercial space sector. The article firstly outlines the challenges in constellation mission planning, including the increasing computation requirements, dynamic constellation capacity, the poor adaptability of node destruction, and the difficulty of establishing target selection rules to satisfy the rapid changes of the mission. In this paper, we propose an autonomous mission planning algorithm based on a dual-core processor which includes a mission receiving mechanism, a satellite on-board conflict adjudication based on a point system, a mission management process, a mission conflict processing, and a mission execution and feedback mechanism. The resource consumption and performance indexes of the on-orbit mission planning algorithm are analyzed and the constellations collaborative mission planning algorithm in- orbit validated by using two remote sensing satellites, and the validation results show that that the autonomous mission planning by fast orbital extrapolation is able to execute the imaging tasks precisely.

To address the issue of insufficient accuracy in the embedded design of traditional sports performance verification devices, this study innovatively proposes an embedded design scheme for a sports performance accuracy verification system based on intelligent vision technology. In terms of hardware configuration, a high-efficiency Advanced RISC Machines (ARM) processor is selected to replace the original bulky computing unit, outdated embedded controllers, and complex independent power modules. This significantly enhances the system’s integration and energy efficiency. On the software level, advanced intelligent vision technology is ingeniously integrated. Through the meticulously constructed reverse vision model design, the system not only achieves fine classification and evaluation of sports performance but also optimizes the execution efficiency of the underlying code, ensuring the standardized and regulated operation of the model. This model is customized for high-precision performance determination and can adapt to the scoring rules of various sports events. To verify the practical effectiveness of this system, simulated application scenarios were constructed, and detailed simulation tests were conducted. The analysis of the experimental data shows that the embedded sports performance verification system based on intelligent vision significantly improves the accuracy and efficiency of performance evaluation. It also demonstrates excellent stability and versatility, thus strongly confirming the effectiveness and innovative value of this design.

In order to explore the application of the emergence effect of swarm intelligence in the formation optimization of unmanned aerial vehicle (UAV), this study conducted an in-depth study on the control strategy of UAV formation by constructing a theoretical model based on the swarm intelligence and carrying out simulation experiments. By simulating the swarm behavior of organisms in nature, this study designed a UAV formation control strategy that can adapt to complex environments and task requirements. The results suggest that compared to the traditional centralized control method, the control strategy based on swarm intelligence emergence shows significant advantages in improving the task execution efficiency and enhancing the robustness of the system. This study not only provides a new theoretical support and technical path for the research and application of UAV formation flight technology, but also opens up a new horizon for the development and interdisciplinary application of swarm intelligence theory.

This paper mainly studies and designs an indoor autonomous tracking vehicle system based on FreeRTOS real-time operating system and UWB ultra-wideband ranging technology. The system integrates advanced wireless positioning technology and embedded control algorithm to achieve intelligent following function with high accuracy, low delay and strong robustness. Firstly, the characteristics and advantages of FreeRTOS are introduced, and how to apply it to the design of the car control system is described to meet the requirements of real-time control and ensure the stable and efficient operation of the system. Secondly, the principle of UWB ranging technology and its application in precise positioning are discussed. Combining the above two key technologies, a vehicle following strategy based on FreeRTOS and UWB ranging information was proposed. Using high-precision ranging capabilities and excellent task control capabilities, the vehicle can track the preset target object in real time and maintain a safe distance. Finally, the effectiveness and feasibility of the designed system are verified by experiments.

The Spring Boot based car rental system proposed in this paper is a B/S based system aimed at meeting the needs of the car rental industry. The development of a car rental system mainly uses Java technology, combined with the latest popular Spring Boot framework. The MySQL database and Eclipse development environment are used. The car rental system includes two roles: user and administrator. Its main functions include administrator: homepage, personal center, user management, vehicle brand management, vehicle information management, vehicle color management, rental order list management, return record management, administrator management, collection management, system management, user: homepage, personal center, vehicle information management, rental order list management, return record management; Functions such as vehicle information, system announcements, personal center, and backend management. Utilizing the Spring Boot framework to quickly build system infrastructure and integrating Spring Security MyBatis and other technologies enable user authentication and authorization, data access, and other functions. A user interface is built through a front-end framework. to achieve interaction between users and the system. The system supports multiple payment methods, including online payment and in store payment, to meet the different payment habits of users. At the same time, core business processes such as order management and vehicle scheduling have been implemented to ensure the integrity of system functions. The test results indicate that all functions of the system meet the design requirements and have high practical value and application prospects.

The study proposes a multimodal fusion framework termed Vision-Motion Multimodal Classification (VMMC), with the objective of addressing the inherent complexities in the classification of driver styles. This framework amalgamates visual and motion data, leveraging the harmonious interplay between the visual modality feature extraction module and the motion feature extraction module, complemented by the integration of a cross-modal attention mechanism, to achieve precise classification of driver driving styles. Through meticulous experimental evaluation, the VMMC framework demonstrates substantial advantages across metrics such as precision, recall, and F1 score, thus validating the superiority of the VMMC framework. These research findings not only provide novel perspectives on the application of multimodal fusion in driver style classification but also offer invaluable insights for a deeper understanding of driving style patterns.

Due to the lack of sufficient exploration of global spatial-temporal characteristics of traffic dynamics, large-scale and longterm vehicle speed prediction problems have not been well solved. To this end, this study presents a 3-dimensional road matrix based deep learning model (3DM-DLM) to embed global similarities of road segments in constructing learning matrix. The global similarity is measured by sampling the correlation of vehicle speed, and the correlation aggregation of adjacent road segments in the matrix is realized by clustering and Z-order curve. The learning matrix is then used to train a deep neural network composed by convolution layers and residual units. We collected traffic speed data in Beijing to validate the 3DM-DLM. The results showed that compared with the baseline model, the prediction accuracy of the proposed model is improved by 8.05 % in the acceptable time, and it also proved the generalization ability of 3DM-DLM in specific cases.

This paper explains the current application status and future development space of unmanned aerial vehicle (UAV) assisted power grid inspection. Based on existing technical research and practical applications, this paper analyzes and prospects from four aspects: channel inspection, tree obstacle inspection, detailed tower inspection, inspection data management and application. The research shows that the UAV-assisted power grid inspection technology has entered a rapid growth stage from the exploratory stage and has a broad application prospect.

Natural disasters often inflict significant damage on communication infrastructure, which plays a crucial role in emergency response operations. With technological advancements, UAV communication network (UAVCN) is now capable of providing sustained communication services. However, existing research typically focuses solely on the objective of maximizing communication coverage, while traditional genetic algorithms are susceptible to converging on local optima. Therefore, this study establishes a multi-objective planning model and employs an improved genetic algorithm for its solution. Initially, a two-stage stochastic planning method is proposed. The first stage determines the optimal layout of UAVCN to maximize communication coverage and throughput, while the second stage generates the optimal paths for UAVCN to minimize rescue time and energy consumption. In solving the two-stage model, an improved genetic algorithm (IGA) is adopted, which combines global search capabilities with rapid convergence. Finally, the 2017 Jiuzhaigou earthquake is selected as a case study to construct and simulate the two-stage model, thereby verifying the effectiveness and feasibility of the model and algorithm, and to obtain the optimal planning scheme.

This paper establishes a population competition model to analyze the competitive relationship between traditional fuel vehicles and new energy vehicles based on the number of production and sales and market share of traditional fuel vehicles and new energy vehicles in the past ten years. of new energy vehicles, GDP, sulfur dioxide, nitrogen oxides, coal consumption, crude oil consumption and other factors on carbon emissions, multiple regression analysis and particles warm algorithm. regression analysis and particle swarm algorithms are established to predict the time of carbon peak and carbon neutralization. The conclusions of this paper are as follows: the rapid development of new energy vehicles and the development of carbon neutralization in the past ten years. The conclusions of this paper are as follows: the rapid development of new energy vehicles has had a great impact on the traditional fuel vehicle market; the development of new energy vehicles has a negative correlation with carbon emissions. The conclusions of this paper are as follows: the rapid development of new energy vehicles has had a great impact on the traditional fuel vehicle market; the development of new energy vehicles has a negative correlation with carbon emissions, and its development can promote the advancement of dual-carbon; and this paper predicts that carbon peak can be reached in 2030, and carbon neutrality can be reached in 2060.

The task of environmental perception is an extremely important part of the automatic driving project, and mainstream schemes are divided into visual perception and radar perception. Visual algorithms have advantages in cost and object detection or image segmentation methods are usually used for perception. We adopt computer vision scheme to collect data and establish Transportation of Wuhan China (TOWC) datasets based on Chinese urban situation. After optimizing the selection of anchor boxes, the Yolov4 model is retrained as the object detector, which is combined with Simple Online and Real-time Tracking (SORT) algorithm and Perspective-n-Point (PnP) monocular algorithm to realize the integration of object detection, tracking. The detection and tracking methods have been evaluated against the Microsoft Common Objects in Context (MSCOCO) datasets and Multiple Object Tracking (MOT) challenge respectively. And the detection performance also has been evaluated again in our TOWC datasets. The evaluation has been conducted in challenging conditions, including occlusion, partial visibility, and under lighting variations with the mean average precision of 97.17% and the real-time speed of 30.2 FPS. The results show that our fusion detection algorithm is not inferior to other most advanced methods on the premise of excellent real-time performance. The developed model is a generic and accurate solution for traffic object detection and tracking, which can be applied to many other fields, such as automatic driving vehicle, traffic sign recognition, anomaly detection, motion analysis, or any other research areas where the traffic environmental perception is in the center of attention.

Currently, with the development of society and advancements in technology, there is an increasing need for elderly and disabled people to improve their quality of life and self-reliance. Under such context, designing an smart wheelchair with easy operation, comprehensive performance, and health monitoring functions has become a significant focus in the market. In this project, we used the inertial sensor MPU150 for wheelchair motion direction and speed control, designed the wheelchair control system into the form of glasses. We also improved the PPG probe and its analog front end, presenting an innovative method to detect the weak pulse wave signal from the user’s head. Furthermore, the vital signs collected by the wheelchair can be shared through the internet for medical staff and family members to check the health conditions of the user in real-time.

This paper explores a new approach to product quality control based on Agent-Based Modeling and Simulation (ABMS). Addressing the limitations of traditional methods in dealing with multi-factor coupling and dynamic changes, this study constructs a multi-agent system model that simulates the impact of factors such as human, machine, material, method, and environment on product quality loss. Utilizing the Netlogo simulation platform, a dynamic analysis is conducted to assess the contribution of each factor to quality loss, offering a fresh perspective on quality control. Additionally, this paper proposes a product structure optimization strategy based on the model, identifying key parameters through sensitivity analysis to provide a scientific basis for performance optimization. This research not only enriches the application of ABMS in the field of quality control but also provides effective tools for quality improvement and structure optimization in actual production, carrying significant theoretical and practical implications.

Due to the adaptive transmission components of the hydraulic torque converter, compact planetary gear transmission mechanism and proportional valve precise control of the multi-clutch control system, AT transmission has good road adaptation, outstanding starting and acceleration characteristics, high power density, comfort shift process and other advantages. It is widely used in military, mining and other vehicles with complex road conditions and requiring high mobility flexibility. The combination of the multi-degree of freedom planetary transmission mechanism and the clutch highlights the power shift advantage of the AT transmission. Improving power performance and shifting comfort is an important design content of military and mining vehicles, which involves the design and matching of parameters from engine to transmission and each transmission part of the vehicle. Based on Simulink, this paper analyzes and studies the dynamic model of vehicle driving through, and examines its application in vehicle driving through.

The Shenzhen-Zhongshan Link, crossing the Pearl River estuary in China, is a world-class cross-sea cluster project featuring “bridge-island-tunnel-undersea interconnection”. It is a major national project and a core transportation infrastructure in the Guangdong-Hong Kong-Macao Greater Bay Area. The steel structure of the bridge and tunnel is large in scale and difficult to construct. After years of technological research and development, the project has established a system for intelligent manufacturing technology of steel structures for bridges and tunnels through theoretical analysis, experimental calibration, equipment development, process testing, and engineering verification, achieving favorable results. The project has proposed a ‘four-line, one-system’ technological framework for the intelligent manufacturing of steel structures in cross-sea cluster engineering projects, encompassing bridges and tunnels. It has developed a hybrid programming technology for complex components based on features and visual recognition using robotics. The project has also overcome key technologies for the high-precision manufacturing of extra-wide and variable-width immersed steel shells made of special materials, as well as three-dimensional measurement and control methods. Additionally, it has developed a full penetration welding joint for the U-ribs of orthotropic steel bridge decks using double-sided submerged arc welding. his has significantly enhanced the level of steel structure manufacturing, leading to the transformation and upgrading of the steel structure manufacturing industry.

In recent years, neural-network-based adaptive dynamic models are commonly used to estimate and control flight dynamics for drones and multi-copters. However, most of them simply use networks to optimize few parameters in control policy. There are still large improvements on the structural layout for robust model-free control applications. Therefore, to achieve a similar intelligence of human is still challenging due to their difference in basic mechanisms and difficulty in network modeling. In this paper, we design a control learning algorithm which combines reinforcement learning with neural networks simplified from human cerebellar motor learning model. The algorithm learns parameters by statistically measuring the performance and analyzing the input-output relationship on real-time episodes. In local linear systems, parameters are learned with respect to a spatial function of environment state and subjective expectation. Compared with other methods using static models, the most obvious advantage of this algorithm is that it can learn complex dynamics of alternative degrees of freedom while the dynamics are difficult to be formulated by equation set. Besides, the algorithm is suitable for individual systems, without prior knowledge about system geometry, centre of gravity as well as installation error, since it learns the dynamic effects directly relevant to the optimal guidance and control behavior in unknown or partially known environments instead. Experiment verifies the algorithm in a practical way. In the experiment, the algorithm is implemented to a quad-copter and it can learn the flight control policy from zero-state without any prior knowledge. The flight quality is tested to be equal to accurate control model at outdoor flight experiences. By repeated experiment, the algorithm is demonstrated to have good robustness to control different physical models and the potential to explore alternative dimensionality.

In order to meet the requirements of real-time and accuracy for unmanned aerial vehicle (UAV) inspection of transmission lines, this paper deeply studies the application of YOLOV3 object detection algorithm in the onboard AI module of UAV inspection. By integrating the target detection candidate region selection and object recognition into one, the YOLOV3 algorithm, combined with multi-scale feature fusion, realizes high accuracy and real-time optimization of target detection and uses residual blocks to solve the problem of model degradation. The test results of transmission line insulators show that the average accuracy of YOLOV3 algorithm can reach 90%. Under the same conditions, the average processing speed of YOLOV3 algorithm is about 3.2 times that of Faster RCNN algorithm.

Over the years, U-Net has become a predominant model in the domain of retinal vessel image segmentation. However, its constrained receptive field and the inherent biases associated with convolutional operations present significant challenges in effectively capturing long-range dependencies. In recent years, although Transformer-based techniques have been integrated into the U-Net architecture to overcome this limitation, the self-attention mechanism inherent in Transformers demands substantial computational resources, thereby increasing computational complexity and the risk of overfitting. To address these challenges, we propose a model that integrates lightweight Transformer and CNN networks, namely MobileViTv2-ResUNet, for precise retinal vessel segmentation. We chose U-Net as the framework for the automated retinal vessel segmentation model. Firstly, in the encoding phase, we introduced MobileViTv2 blocks to replace traditional convolutional modules for feature extraction. Subsequently, inverted residuals are employed within the encoding phase to perform downsampling operations, thereby reducing computational complexity while enhancing the network’s representation and generalization capabilities. Additionally, an ASPP module is incorporated between the encoder and decoder to effectively fuse feature information from different scales. Finally, in the decoding phase, we integrate our designed LeakyRes module to prevent the occurrence of the “neuron death” phenomenon, thereby improving the accuracy of retinal vessel segmentation. We validated our MobileViTv2-ResUNet on the public datasets HRF and STARE. Experimental results demonstrate that our MobileViTv2-ResUNet outperforms most existing state-ofthe-art algorithms, significantly enhancing vessel segmentation methods, particularly for images with anomalies, bifurcations, and microvessel segmentation challenges.

Recently, the vision transformer (ViT) model of deep learning has achieved surprising performance in the field of computer vision and has been widely used in IoT edge devices. However, the training of ViT models requires a large amount of data and computing resources, which is a challenge for resource-constrained edge IoT devices. To solve the above problems, this paper proposed a lightweight ViT method based on transfer learning. The primary concept of this method is to train large-scale ViT models in the cloud (CloudViT) and deploying small-scale ViT models at the edge (EdgeViT). Firstly, through the method of transfer learning, some underlying parameters of CloudViT were utilized to construct EdgeViT. The purpose is to enable EdgeViT to learn from CloudViT, acquiring knowledge and improving its performance. Secondly, adding a randomly initialized LayerNorm layer before the MLPHead during the training process of EdgeViT, it can improve further model performance. Finally, Experiment results demonstrated that EdgeViT could achieve 91.3% of CloudViT’s performance with only half the parameters and floating-point operations (FLOPs). Moreover, finetuning EdgeViT with a 60% reduction in training time still allows it to achieve 81.3% of CloudViT’s performance. Relevant conclusions can provide technical support for the proposed method.

In the field of deep learning, convolutional neural networks and transformer architecture have achieved considerable success. This paper combines the advantages of these two architectures to achieve the encoding of multi-scale spatial features in an image. This is done by using convolutional kernels of different sizes for convolutional operations at each stage. This allows the model to obtain markers with rich and diverse features. The self-attention mechanism is then used to further improve the feature representation and introduce residual links. The experimental results demonstrate that the proposed model exhibits robust performance on the CIFAR-100 and CIFAR-10 datasets, with comparable performance and fewer parameters compared to traditional CNN models.

With the rapid development of mobile Internet and the increasing maturity of virtual reality technology, there is an increasing demand for immersive virtual reality experiences on Web platforms. Therefore, this paper aims to explore the development of virtual reality technology on the Web side, and WebGL technology is adopted to realize this goal. The paper details several commonly used WebGL development frameworks and evaluates them in various aspects. It also proposes a generic framework design for Web3D modeling and a specific construction method for realizing 3D virtual scenes. Finally, the paper discusses the performance optimization techniques for Web3D from the perspectives of modeling, loading, rendering, and memory optimization. In conclusion, this paper explores WebGL-based virtual reality on web platforms, presenting a general Web3D modeling framework and performance optimization techniques for efficient and responsive 3D virtual scenes.

In order to ensure the safety of navigation of ships at the intersection of traffic flow, the design scheme of the cautionary area in the area of dense traffic flow should be evaluated for risk. This paper constructs a navigation safety evaluation index system for the cautionary area from four aspects: natural environment, traffic environment, ship factors, and management factors. Making full use of the advantages of COWA operator and improved topological cloud theory, the safety level of navigation in the cautionary area is evaluated based on the improved topological cloud model, and the safety evaluation of navigation before and after setting up the cautionary area is carried out with the example of the No.1 cautionary area in Qinzhou Bay. The example proves that the evaluation model is correct and practical, and can be used for the safety evaluation of navigation in the cautionary area.

As an effective tool for knowledge representation and uncertainty reasoning, Bayesian networks (BNs) are widely used in various fields. However, learning the structure of BN is an NP-hard problem. It is impractical to rely solely on the experience and knowledge of domain experts to build BN. Data-driven learning of BN has become a necessity. For the learning of a BN structure with data containing continuous variables, the typical method is to discretize the data or assume that the data follows the Gaussian distribution, and then apply the traditional BN structure learning methods to discover the causal relationship. The discretization inevitably leads to the loss of valuable information of the data. Realworld data sometimes may not follow the Gaussian distribution, which can cause deviation in causality. In this paper, a new constraint-based BN learning method for continuous variables is proposed for BN structure learning. Mutual information and conditional mutual information are derived by a non-parametric kernel density estimation (KDE). The correlation between any two nodes can be determined without assumptions. As new conditional independence tests, they are used in the max-min parents and children (MMPC) algorithm, which is a typical constraint-based method. We compare the proposed method with traditional BN methods using well-known benchmark networks. Synthetic continuous data are generated by linear structural equations. The experimental results show that our method has a good performance. It can be used as an effective BN structure learning method for continuous variables.

Considering the uncertainty of demand and transport time in the actual transport process, a fuzzy nonlinear planning model for multimodal transport considering hybrid time window constraints is established with the optimisation objective of the integrated operating cost of transport cost, transit cost, time penalty cost and carbon emission cost. The optimisation model containing uncertain variables is transformed into a mathematical model of deterministic form through the fuzzy opportunity constraint theory, and the hybrid simulated annealing algorithm combining genetic algorithm and simulated annealing algorithm is used for the solution, and the validity of the model and algorithm, as well as the effect of time window on the results, are verified by changing the time window constraints.

The worldwide incidence of breast cancer continues to increase annually. Accurate classification of breast cancer subtypes is crucial for improving treatment precision. However, the diagnosis of breast cancer is time-consuming and labor-intensive. Although the introduction of Computer-Aided Diagnosis (CAD) systems can assist doctors in diagnosing breast cancer more effectively, establishing high-quality medical image datasets is challenging, significantly hampering the development of CAD systems. This paper proposes a meta-learning method suitable for the multi-classification of breast cancer histopathological images. This method effectively addresses the issue of limited data by adopting the concept of transfer learning and incorporates group convolution within the model to enhance efficiency and feature extraction capability. Experimental results demonstrate that this method achieves a classification accuracy of 96.30% on the BreakHis dataset across eight categories, surpassing state-of-the-art methods. Furthermore, we validated the generalization performance of this method on another publicly available dataset, Kather-CRC-2016.

To improve the accuracy of identifying non-conforming parts and improve production efficiency, this paper proposes an intelligent identification method for micro defects in metal parts based on the YOLOv5 algorithm. This method first preprocesses the collected images and then extracts part features by enhancing fixed features, extracting image defect feature points, and training the YOLO network. Then, based on the YOLOv5 algorithm, an intelligent recognition model is established to achieve intelligent recognition of micro defects in parts. The experimental results show that the detection accuracy of this method is higher than 93.9%, with an average detection time of 2.57ms, which is better than the comparison method and has an ideal detection effect.

Cells are capable of conveying rich information about their microenvironment. Automated cell nucleus segmentation technology not only alleviates the workload of pathologists but also provides accurate microenvironment data applied to biological and clinical research. Existing deep learning models show excellent performance given a large volume of labeled data, they still usually require some manual labeling for domain-specific training when faced with data from unknown fields. Unfortunately, obtaining precise annotations of histopathological images is an extremely challenging task that is not only highly dependent on expert knowledge but also takes a significant amount of time. In this study, we aim to develop a general cell nucleus model for segmentation that requires a smaller amount of data and has broader applicability. To this end, we designed a Dual Encoder U-net model based on Meta Transfer Learning (MTL), which can learn effectively with fewer training samples. Our network employs a Meta Transfer Learning optimization algorithm and introduces a feature fusion module to integrate deep and shallow feature information. Experimental results show that our proposed MTLDEUnet model not only enhances the generalization ability of the model but also achieves performance comparable to the current highly advanced models trained with a limited sample set.

To address the challenges in gastric cancer pathological images, such as varying sizes and shapes of lesion regions as well as blurry boundaries, we propose an enhanced U-Net architecture segmentation algorithm based on an affine crossattention mechanism. Specifically, we introduce affine transformation modules into the up-sampling and down-sampling stages of the U-Net, replacing adjacent convolutional blocks to better capture variations in shape and size. Additionally, a cross-attention module is incorporated in the bridging phase to enhance feature utilization and mitigate mis-segmentation of healthy tissues. In contrast to the conventional U-Net, our algorithm demonstrates notable enhancements in terms of 8.21%, 6.87%, and 5.57% in Dice coefficient, Intersection over Union (IoU), and Accuracy (ACC), respectively. The effectiveness of our introduced modules is reinforced through ablation investigations. The segmentation performance of lesion regions in gastric cancer pathological images is augmented by the proposed algorithm, as shown by the experimental results, effectively reducing the false-positive rate in image diagnosis

Imbalanced classification tasks, prevalent in industrial fault diagnosis, network intrusion detection, and disease identification, pose challenges due to limited samples, intricate inter-class relationships, and overlapping boundaries. Addressing these issues, this paper introduces CSDSResNet, a novel cost-sensitive dual-stream residual network for multi-class imbalanced data classification. To address limited samples in the minority class of imbalanced datasets, we introduce a dual-stream residual network for enhanced feature extraction. Additionally, a unique cost-sensitive loss function is designed to navigate complexities arising from imbalanced inter-class sample quantities and overlapping boundaries. Emphasizing the minority class and challenging classes with high inter-class similarity, this loss function significantly improves the model’s classification ability. Evaluation of ‘DryBeans’ dataset reveals CSDSResNet’s superiority, surpassing state-of-the-art methods by 2.9% in macro_F1score, with the highest precision in minority class recognition.

In the operation of train pick-up and delivery, there are many problems with human inspection, resulting in potential safety hazards. Considering the wide variety of railway trains, the appearance of the train tarpaulin is selected as the research object, and the damage of the tarpaulin is automatically detected by combining object detection and image segmentation techniques. When using object detection to detect damage, the edge damage detection effect is not ideal, so the image segmentation algorithm is combined with the object detection algorithm, while the traditional segmentation algorithm is not ideal for small damage and narrow and long damage segmentation effects, resulting in low detection accuracy. Therefore, a segmentation algorithm DMV3 is designed, and the DMV3 algorithm is feasible through experiments, which solves the problems of inobvious segmentation and inaccurate detection. Compared with other lightweight segmentation algorithms, it also has great advantages.

Particle swarm optimization (termed as PSO) is a meta-heuristic searching technique that is employed to tackle a variety of optimization issues in the real world, and it has become one of the hotspots for scholars’ research. PSO own the advantages of few adjustable parameters, fast convergence, and simple search process, etc. However, it still has some intrinsic defects, including the disadvantages of easy to plunge into the local optima and low convergence precision in late evolution stage. To solve the issues mentioned above, an adaptive learning scheme-based particle swarm algorithm (termed as APSO) is proposed. To begin with, the global and local search are balanced by improving parameters inertia weight, personal and social acceleration coefficients, respectively. After that, different adaptive position update mechanism is used according to different evolutionary states so as to achieve better convergence. At length, conducted experiments based on 12 classical test functions reveal the merits of the proposed APSO. The simulation experiments and the evolution curves show that APSO markedly outperforms the canonical particle swarm optimization in accordance with stability, convergence rate and solution precision.

In order to avoid the safety hazards of fishing vessels dispersing and crossing the precautionary area in a disorderly manner, reducing collisions between commercial and fishing vessels and ensuring the safe and efficient operation of neighboring ports, it is crucial to select the area where fishing vessels are concentrated to cross the large precautionary area. Based on the advantages and disadvantages of the particle swarm algorithm and the path planning principles outlined in the General Provisions on Ships’ Routeing, we conducted a site selection study focused on the waters within the large cordoned-off area where fishing vessels frequently cross. As a practical example, we planned a recommended route for fishing vessels navigating the densely trafficked area in the southeast of Shidao. The recommended route for fishing vessels described in this plan is the optimal crossing of waters under the condition of non-closed fishing season, which was verified to be feasible by examples and announced on April 9, 2024, for implementation.

In this work, a semi-analytical model of shock response spectrum (SRS) in rectangular plates under impulse loads is carried out. Firstly, the modal analysis technique is employed to solve the acceleration response of rectangular ReissnerMindlin plates with simply-supported edges. Then, semi-analytical expression of corresponding SRS is obtained after introducing the convolution integrals. With the semi-analytical model, the characteristics of SRS with input variable parameters, such as excitation load and dynamic parameters of plates, are analyzed systematically. Subsequently, a finite element model (FEM) of rectangular Reissner-Mindlin plate is developed to study and verify the regulation effect of excitation load (peak and duration) on the characteristics of SRS. The simulation results indicate that the response accelerations and corresponding SRS of plate agree well with the regulations from semi-analytical model. Lastly, rectangular Reissner-Mindlin plates with various variable parameters are simulated with FEM model to further explore the regulation effects of SRS, and results indicate characteristics of SRS are highly dependent on each input variable (excitation load, dynamic characteristic, response position and boundary condition of plate).

In this article, a lightweight face recognition algorithm is constructed, which is based on the improved MobileFaceNet. To improve recognition accuracy and meet real-time requirements under the premise of ensuring a lightweight model, the inverse residual network of the ECA-Net network and H-swish activation function is designed. ECA-Net network enhances network cross-channel learning ability to improve algorithm accuracy and replaces ECA-Net network activation function with H-swish to enhance model device applicability.

Small-scale pedestrian detection is a major challenge due to limited pixel resolution and insufficient distinguishing features, frequently resulting in incorrect or missed detections. To address it, this paper proposed a global context-aware and attention mechanism algorithm for small-scale pedestrian detection. Firstly, considering the problem of small-scale pedestrian features gradually decreasing with network depth, we leverage the advantage of Transformers in capturing longrange dependencies. This allows us to design a global context information module that can retain a large number of smallscale pedestrian features. Then, considering the issue of small-scale pedestrian features easily being confused with background information, a Coordinate and Channel Attention Module (CCAM) is proposed. Coordinate attention can capture direction-aware and position-sensitive information, which helps the model to locate and recognize objects of interest more accurately. Channel Attention can effectively enhance small-scale pedestrian features and suppressing background information. Experimental results on the CrowdHuman dataset fully demonstrate that the proposed method can significantly improve the detection ability for small-scale pedestrian.

The fatigue detection method based on neural network has been widely applied in the field of land transportation. The accuracy varies by different methods and solutions. Disadvantages of current widely applied methods for fatigue detection have been exposed by rational review, e.g., cost-effective, potable implementation issues. Cascade width learning based detection method with outstanding features is a sound solution to overcome shortcomings of other traditional detection methods. Therefore, it is of great significance and value to research this method and present its academic value and applicable values. This paper introduces a cascade width learning based fatigue detection method. The paper looks forward to development of cascade width learning based fatigue detection method, both academic and applicable values of proposed method have been presented by rational evaluation based on rational discussion on experimental results, the results prove that proposed method has achieved 94.9% accuracy in 52.43ms, which suggests more accuracy and fast speed than other common methods including: combinations of LBP and SVM, ASM and Fuzzy, CNN and PERCLO.

Our previous research showed that emulating dendritic neuron structure effectively addresses orientation detection challenges in learning tasks, reducing both learning time and costs compared to alternative neural network approaches. Our simulation model incorporates an On-Off Response mechanism in bipolar cells (BCs) and horizontal cells (HCs) for processing grayscale input images. To overcome limitations with single-channel input, we introduce color-selective cells. By integrating these cells, we enhance and select outputs of local orientation detection dendritic neurons, generating specific feature maps. These feature maps are generated using a biologically-inspired model that mimics the mechanism of color-selective cells. Additionally, global neurons are used to capture overall image features by aggregating outputs from local dendritic neurons. Our system utilizes backpropagation to update parameters of local orientation detection dendritic neurons. Furthermore, we integrate a learnable orientation detection neural network after the dendritic neuron stage.

Taking emergency materials scheduling as the research object, this paper puts forward the problem that deadlock is easy to occur in the scheduling process of emergency materials after various major events. Since Banker algorithm can effectively avoid deadlock in resource scheduling process, an emergency materials scheduling system based on Banker algorithm is designed.Through the design and implementation of the module functions of system management, emergency materials management and scheduling, and information sharing, the efficiency of emergency materials scheduling is improved, the real-time monitoring of the whole process of emergency resource application and scheduling is realized, and the information level of emergency resource management is enhanced.

The core is the first-hand physical information that can directly reflect the formation conditions. Core data preservation become one of the concerns of scientific researchers, with the development of core image high resolution acquisition and processing technology and application of increasingly mature, 3 d digital core technology become an important technical means of complete preservation core library, greatly reduced the weathering, broken and sample selection natural or artificial reasons such as data distortion. This study is the first to introduce this technology into the field of live fault exploration, which can provide new methods and beneficial exploration for the core research in the field of geology.

In the era of digitization and widespread access to information, in order to create a good academic atmosphere, promote academic integrity, and ensure the authenticity and originality of electronic homework, a plagiarism checking system for college students has been designed. The system uses a combination of natural language processing algorithm SimHash and electronic document attributes to detect similarities between submitted tasks. By analyzing document attribute characteristics and document content similarity, the system can accurately and efficiently identify potential plagiarism behaviors. This system is developed using the SSM framework, and the MariaDB relational database management system stores industry data. The system allows teachers to create courses, assign homework, and check homework plagiarism. Students can check homework online, submit homework, and view homework grades. The application results of this electronic homework plagiarism checking software in the course of Database Principles and Applications show that teachers can quickly learn about student homework plagiarism, objectively evaluate the quality of student homework completion, and also play a warning role for students. It improves the learning atmosphere, promotes the healthy growth of students, and adapts to the development trend of digital teaching assistance software in the era of artificial intelligence.

Chinese excellent traditional culture is the source of Chinese national spirit. As the core value system of Chinese traditional culture, Chinese medicine is a treasure of the Chinese nation. The spread of Chinese medicine is an important way for China’s values to be known and accepted globally. In the past, the means of communication were not rich without many people taking part in it, so it is of great practical significance to build a multi-modal interaction mechanism. With the indepth development and important influence of Artificial Intelligence technology in all aspects of social life, the characteristics of high interactivity, popularity and accuracy of artificial intelligence technology provide new perspectives for the international spread of Chinese medicine and contribute to the construction of its multi-modal interaction mechanism.

With the development of power electronics technology, converters are increasingly widely used in industry and daily life. Pulse Width Modulation (PWM) technology, due to its high efficiency and flexibility, occupies an important position in the control of converters. This paper explores the application of PWM in converters through simulation research, analyzes the impact of PWM control strategies on the performance of converters, and provides beneficial references and insights for further research and experiments.

With the continuous progress of society and the rapid development of technology, emerging technologies such as artificial intelligence are becoming increasingly prevalent in daily life. In particular, in the field of human pose recognition, traditional solutions have mostly relied on high-performance servers or specialized hardware, which have limitations such as high cost, poor real-time performance, and low convenience. Therefore, developing a low-cost, high-performance human pose recognition system is of great significance for promoting technological progress and meeting people's needs for intelligent living. TensorFlow.js is an open-source machine learning library that makes it possible for AI models to run in Web browser environments. On the basis of TensorFlow.js technology, combined with the front-endVue.js framework, an online real-time human pose recognition system was developed on the Web browser end. By calling the PoseNet model and utilizing the CNN model to optimize the overall learning performance, the COCO key points are defined and recognized, and the pose recognition model is deployed on the Web browser end, lowering the system's usage threshold and improving user access efficiency. At the same time, it ensures a relatively high recognition accuracy, reducing dependence on server resources, and achieving lightweight, low-cost real-time pose recognition detection and analysis function.

The main content of this study is the overall design and key structure design of a certain type of 2t forklift, including the determination of the main performance parameters and performance requirements of the forklift, the selection of the main system assembly scheme and the design calculation of its key structure, and the overall layout design of the above assembly; in order to ensure that the designed electric forklift has good working performance, it is necessary to verify and evaluate the main performance and structural strength of the vehicle. In this paper, we verify whether the design of the key structure of the electric forklift meets the industry standards by comparing the differences between the theoretical calculation results and the simulation results.

Currently, there are several issues with predicting the situational awareness log traffic of various security devices monitored within the security service networks of operators and internet enterprises. Firstly, most existing prediction models lack the capability to extract spatio-temporal global information from message sequences and tend to have a high proportion of ineffective feature information. Secondly, in the actual production network environment, there is a strong trend correlation between the first-order differences of situational alarm log traffic of each device at consecutive time points and the network delay volatility of each device’s received logs. However, most existing traffic prediction methods fail to consider this aspect, resulting in lower prediction accuracy of network situational traffic. To address these issues, an improved spatio-temporal transform network model introducing Volatility Evaluation (VE-STTN) is proposed for predicting situational log traffic and handling data. The VE-STTN model not only introduces a dynamic pooling layer into the existing STTN network, reducing the extraction of ineffective features from log traffic data in spatio-temporal features and achieving key information aggregation but also enhancing the model’s learning performance. Particularly noteworthy is the innovative introduction of the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model data processing module in the VE-STTN model to calculate the forecasted future delay volatility from the improved STTN predictions and use it to optimize and adjust the predicted situational traffic results. Experimental results demonstrate that this approach improves prediction accuracy and robustness

Camouflage target detection is a difficult problem in the field of target detection. The practice of camouflage target detection based on depth learning shows that the scale proportion of the targets in the image is one of the important factors that affect the detection performance. Based on the region segmentation and texture analysis, two schemes of camouflage target filtering are proposed in this paper. Experiments show that the proposed schemes improve the integrity of the target filtering significantly, it lays the technical foundation for improving the accuracy of deep recognition in the later stage.

In this paper, an efficient object detection method YOLO-Ti is proposed to detect tiny facial markers. Our study is driven by the practical requirements of 3D face modeling, requiring the incorporation of as many facial features as possible for reference. This research can even provide information for facial expression recognition and joint deformation. To achieve this, we first present a feature fusion module called Cross-BiFPN, which incorporates additional crossconnecting branches between different network layers to utilize low-level features more effectively. Secondly, we add a high-resolution detection head and attention module to the YOLOv8 model to improve the ability of detecting tiny objects, while at the same time ensuring the lightweight detection model by reducing redundant network layers. Thirdly, we collect a dataset of facial markers with an average size much smaller than publicly available small object datasets. Ablation studies and comparison experiments are conducted to evaluate the performance of our approach. Compared with the baseline YOLOv8 model, YOLO-Ti shows a 30.4% improvement in mAP50 while reducing model parameters by 65.1%. The automatic feature extraction provided by our model facilitates the construction of digital humans, providing significant savings in manpower and time for modelers.

In this paper we consider a nonlinear Bertrand game model and investigate the dynamical properties. It is assumed that the two competitors follow linear cost functions and quadratic emission charges. We construct a two-dimensional system for the model first. Then we examine the equilibrium points and their stability conditions. The chaotic properties are presented numerically via bifurcation diagrams, maximum Lyapunov exponent, sensitive dependence on initial conditions and strange attractors. Finally, we control the chaotic behavior by time-delay feedback method.

This paper evaluates four AI-based 3D reconstruction algorithms—TripoSR, Meshy, Instant, and One-2-34—alongside the traditional AR-Code method (which uses scanning for model point cloud reconstruction). By comparing the integrity ratio from point cloud data, depth information from depth maps, and perceived similarity from model dimensions, we identify the AI algorithm that best reconstructs 3D models. This research aligns with the “Opinions on Promoting the Implementation of the National Cultural Digitization Strategy” and provides actionable cases, methods, and strategies. It has practical applications in cultural heritage preservation, product design, and the digital transformation of smart cities, offering resource efficiency and enhanced user experiences in digital environments.

At present, vehicle routing optimization has become the key to improving logistics efficiency and reducing costs. This article proposes an improved ant colony algorithm to address the limitations of traditional ant colony algorithms in the optimal path problem for vehicles. The core of this study is to improve the pheromone update model of ant colony algorithm and validate it by constructing an experimental environment. The improved ant colony algorithm proposed in this article has significant performance improvements in solving vehicle path optimization problems, and is feasible and superior in practical applications, especially in terms of search efficiency. This algorithm provides a new perspective for future research directions.

Group delay is an important parameter of navigation receiverability. When the navigation signal passes through the RF front-end, it will generate signal delay, which affects the positioning accuracy. The receiver RF front-end has the function of converting RF signals into IF signals, which makes the measurement of its group delay characteristics difficult. This paper firstly introduces the definition and concept of group delay, and from the test mechanism, based on the comb wave generator, it carries out the research on the group delay measurement method of navigation RF front-end, and carries out the experimental verification. The results show that the group delay measurement method based on the comb wave generator has a reliable calibration method and can stably measure the group delay characteristics of the navigation RF front-end.

Federated learning has more flexible data ownership participants, therefore its data (sample feature vector or label) is more likely to be changed, and it is more vulnerable to data poisoning by malicious users, resulting in the final global model not getting the expected effect. This paper focuses on this data poisoning defense problem and applies the traditional centralized machine learning pruning optimization method to each client of federated learning. Each client needs to execute before each global iteration. Pruning optimization algorithm to remove abnormal data. The experimental results indicate that when the discrepancy between abnormal and normal samples is significant, the pruning optimization algorithm effectively eliminates the outliers, thereby minimizing their impact on the final federated learning model.

In this paper, an iterative method is introduced to approximate the solution of nonlinear Volterra integral equation with piecewise intervals. Firstly, we transform the nonlinear integral equation into a system of nonlinear algebraic equations and then proceed to solve these equations. We consider certain conditions of the integral equation and study the convergence analysis of iterative method. A numerical example is provided to confirm the reliability and accuracy of the proposed method.

In this paper, a new method for bridge disease image segmentation is introduced, in which the data set includes exp_rebar, breakage, patch and joint. The proposed method uses the YOLOv8 model to partition the region of disease interest, which serves as the cue input of the Segment Anything Model (SAM) and the high-quality HQ-SAM pre-trained large model, and performs automatic and accurate segmentation based on this. In this study, three evaluation indexes including accuracy, recall rate and F1 score were used to quantify the accuracy of segmentation results of YOLOv8, YOLOv8+SAM and YOLOv8+HQ-SAM models. The results show that the SAM model performs better than the other two models, showing higher segmentation accuracy and overall performance. Although HQ-SAM is improved by SAM, the more complex network architecture did not achieve the expected gain on the dataset in this paper. The YOLOv8+SAM model proposed in this paper provides a new technical direction for the intelligent recognition of bridge diseases.

In recent, machine learning techniques have been employed to predict various diseases such as lung, blood, liver, heart, etc. As the heart is considered one of the human body’s significant organs, this research proposes a comparative analysis of heart disease failure using various machine learning algorithms. Since the ECG and EEG are primary sources for analyzing heart performance, the lipid profile also provides information to determine good and bad cholesterol levels. Although this research proposes a comparative analysis of Heart Failure (HF) and other Cardiovascular diseases (CVDs), it assists in finding a better way to predict the root cause of HF and CVDs. The machine learning models are employed and trained, such as the Support Vector Classifier (SVC), Random Forest Classifier (RF), Logistic Regression (LR), Decision Tree Classifier (DT), and K-Nearest Neighbors Classifier (KNN) on the real-time dataset from Kaggle to predict and classify heart disease patients. The experimental set-up outcomes show that the Logistic Regression (LR) classifier has proven the best accuracy (88.00%) among all other machine learning classifiers. Our research results significantly contribute to predicting (HF) and nurturing advancements in AI-powered tools for improved heart failure (HF) prediction and patient care.

Accurate segmentation of insulators in power systems is crucial for localizing and detecting insulator defects, ensuring the safe and efficient transmission of electricity. However, current insulator segmentation models suffer from high misclassification rates and low segmentation accuracy when segmenting aerial insulator images. To address this issue and achieve precise segmentation of insulator images, we propose an Insulator image Segmentation with Mamba-based U-Net (ISMU-Net). Firstly, an enhanced feature extraction block, grounded in visual state modeling, is devised to replace the convolutional blocks of U-Net, enabling comprehensive extraction of insulator image information. Secondly, to mitigate information loss at skip connections and harness the underlying network information, we integrate the attention mechanism from SE-Net into the feature extraction block, optimizing feature fusion at skip connections. Experimental results on a collected dataset of aerial insulator images reveal that ISMU-Net achieves a Precision (Pre) of 92.7%, Recall (Rec) of 91.7%, and F-Measure (F1) of 94.8%. Moreover, ISMU-Net demonstrates strong generalization capabilities across diverse backgrounds, thereby validating its effectiveness in enhancing the accuracy and robustness of insulator segmentation.

Based on the combination of compressed sensing principle and DNA, a more effective and secure new method for image augmentation is proposed in this paper. After the original image is compressed by the measurement matrix, the DNA coding rule is used to perform the rule operation with the chaotic sequence generated by the Logistic chaotic system. This step is to perform pixel diffusion on the image, and use the Lorenz chaotic system to construct the index sequence to index the diffused image. Scrambling to obtain an encrypted image, and then decrypting it, using the principle of compressed sensing to process the decrypted image, and through the recovery algorithm (this paper uses the OMP reconstruction algorithm), the original image is successfully restored.

Sea sky-line detection is crucial for unmanned vessel attitude estimation and maritime surveillance target detection to reduce computational complexity. Many existing seas sky-line detection algorithms mainly extract sea sky-lines through edge detection, but these methods are less robust and susceptible to water surface ripple interference. In this paper, we propose a sea sky-line detection algorithm based on watershed segmentation, which uses the local binarization method to locate the sea sky-line region, segments the connectivity region by watershed algorithm, selects the straight-line pixel points, and accurately calculates the slope and intercept of the sea sky-line using the LSD and RANSAC algorithms. The experimental results show that the algorithm can accurately detect horizontal lines with less error and outperforms the other five advanced algorithms.

Accurate power time series prediction is crucial for stable operation and optimized scheduling of power grids rapidly developing and integrating increasing renewable energy sources. Traditional prediction models often neglect the spatiotemporal characteristics of power grid data, resulting in inadequate accuracy. To address this, we propose an enhanced dynamic and static graph attention network model for power time series prediction. Experimental results on the Solar Energy and Electricity datasets demonstrate superior fitness values of 99.01 and 99.88 after 30 and 16 iterations, respectively. The model achieves an RMSE value of 2.14% within 100 seconds on the Solar Energy dataset and a CORR value of 0.982 after 30 cycles. In practical application, the method consistently exhibits a low RSE value (within a fluctuation range from 0.021 to 0.035) as the Layer parameter increases. The proposed method offers high prediction accuracy, providing valuable insights for power system management, operation, and scheduling, thereby enhancing the safety, stability, and economic operation of power systems.

Geographic entity relationship extraction is a vital task in the field of Geographic Information Science (GIS) and Natural Language Processing (NLP) that aims to extract relationships between geographic entities from text. The current field of geographic entity relationship extraction lacks a public corpus and the existing models face limitations in addressing long-distance relationship modeling and relationship diversity. Therefore, in this study, we first construct the GeoRelCorpus, a corpus encompassing a wide range of geographic entities and relationships. It is based on the Encyclopedia of China Geography branch, and supplemented with the tagging information of OpenStreetMap. The primary objective is to facilitate the training and evaluation of geographic entity relationship extraction models. In terms of model design, we propose an enhanced CasRel model that integrates a multi-scale feature extraction module, combining IDCNN, BiLSTM, and SENet components to improve feature extraction capability and extraction accuracy. Finally, experiments are conducted on the Baidu entity-relationship extraction dataset and GeoRelCorpus. The results demonstrate a notable enhancement in the F1 value achieved by our improved model, thus confirming its effectiveness.

Low-level wind shear is an important weather phenomenon that affects flight safety, and understanding the horizontal scale of low-level wind shear is highly beneficial for wind shear-related research and flight training. Considering the lack of statistics and research in this area in China, this paper proposes a calibration method for the horizontal scale of low-level wind shear based on the F-factor. First, the vertical wind speed is calculated from various flight parameters recorded in the Quick Access Recorder (QAR) data, and then the F-factor is obtained. Second, the wind shear danger period for triggering reactive wind shear warning segments is determined based on the threshold value (±0.105) proposed by the FAA. Third, the horizontal scale of low-level wind shear is estimated by calculating the distance that the aircraft flies during the danger period of the wind shear. Finally, the QAR data from 147 flights triggering reactive wind shear warnings are statistically analyzed, and the mean value of the common horizontal scale of low-altitude wind shear was estimated to be approximately 517 m, with a standard deviation of approximately 135 m.

Rock mass fractures are one of the main factors leading to slope instability, and the detection of rock mass fractures can predict slope instability to a certain extent. The rapid development of deep learning has provided a low-cost and efficient method for the detection of rock mass fractures. This paper employs the DenseNet121 model and the InceptionV2 model for the detection of rock mass fractures, and improves the models by incorporating an attention mechanism. The dataset consists of rock masses with fractures from various regions to enhance the model’s applicability in different scenarios. Experiments have revealed that the InceptionV2 family of models exhibits overall better performance than the DenseNet121 family of models. Among them, the InceptionV2-ECA model performs the best with an F1 score of 0.9850 and an accuracy rate of 98.73%. Compared to the original InceptionV2 model, the accuracy rate has increased by 9.57%.

To solve the problem of short-run power load forecasting, this article proposes a model using particle swarm optimization (PSO) to adjust the parameters of the backpropagation (BP) neural network, namely the PSO-BP model. Based on this, the GPSO-BP-NN short-term power load forecasting model is constructed. For the sake of verifying the performance of GPSO-BP-NN, actual data from a certain region in China is selected for experimentation. In view of the analysis of the fitness function outcome, by comparing the prediction results of GPSO-BP-NN, PSO-BP-NN, and BPNN models, it is found that the mean absolute error of the GPSO-BP-NN model is 2.21%, which is lower than the 2.39% of the PSO-BP-NN and the 3.53% of the BP-NN. Through the analysis of prediction accuracy, algorithm comparison, and time cost, GPSO-BP-NN is superior to the other two prediction models, proving the efficiency of the improved algorithm

In paper, according to the concrete compression damage constitutive relationship of energy loss, a cylindrical reinforced concrete shear wall with a rectangular section was simulated using ABAQUS software. The skeleton curve for structural ultimate bearing capacity was drawn under the theoretical condition, and the comparative analysis of fitting was carried out, consequently the macroscopic approximate relationship between material damage and structural failure was established, which also proved the rationality of the hypothesis. The advantages of this method are fewer parameters, simple and practical, and high precision, which can provide another solution for the simulation analysis of compressive damage of concrete structures.

In order to explore the cutting force, cutting temperature and cutting parameters and other key factors in the high-speed dry hobbing process between the influence of the relationship and change rule. In this paper, based on the principle of hobbing combined with cutting mechanics and heat transfer theory, the cutting force and cutting temperature during highspeed dry hobbing are modeled, and the relationship between the average cutting force and the feed, cutting speed and back-eating amount is derived from the results of finite-element simulation, and the parametric equations of the tangential force and the radial force empirically are derived from the coordinate transformation and input empirical values, and the empirical parametric regression equations of the axial force are derived, respectively. The method of calculating cutting force and cutting temperature under high-speed dry hobbing conditions is derived, respectively. This research provides a theoretical basis and experimental basis for further exploring the intrinsic mechanism of high-speed dry hobbing, including cutting force, cutting temperature and machining dynamics, etc., so that the traditional hobbing cutting method can be developed in the direction of more energy-saving, high-efficiency and precision.

Traditional algorithms have advantages such as interpretability and portability in pose estimation task. However, in complex background environments, traditional algorithms suffer from poor adaptability and detection errors. When dealing with complex scenes or small targets, CNN-based algorithms exhibit superior accuracy compared with traditional algorithms. However, CNN-based algorithms of pose estimation cannot be further developed on mobile terminals due to the large number of model parameters. To address this problem, this paper proposes the DBayC algorithm. First, the LBN (Limb Behavior Network) module is designed based on the CNN (convolutional neural network) algorithm to achieve the semantic segmentation effect on the human body. Then, the node annotation of human body is performed on the semantic segmentation results from LBN module to form graph-structured data. Finally, Bayesian formula is used to perform conditional probability analysis on the nodes in the graph, and the motion trajectories between nodes are analyzed, thereby achieving pose estimation and behavior analysis. Through the training of two data sets Hi-Eve and PoseTrack2017, and comparison with some SOTAs (state of the art) models. The experimental results show that under Hi-Eve data, DBayC achieved an accuracy of 79.2%, which is 3.8% higher than HRNetV2. Under the PoseTrack2017 data set, the DBayC algorithm achieved an accuracy of 78.6%, 6.9% higher than HRNetV2. It can be concluded that not only the accuracy of the DBayC algorithm has been improved, but the portability of the algorithm has also been improved, so the DBayC algorithm has certain use value.

To reduce the loss of labor force and material resources caused by bird activities at the substations and address the problems of high detection difficulty for small bird targets within complex backgrounds, a bird target detection method based on feature selection and attention mechanism is proposed. Specifically, the feature selection module (FSM) is initially used to extract key information from multi-level features. In order to enable the model concentrating on crucial channels and spatial positions related to the small target, coordinate attention is further embedded in the improved bidirectional feature pyramid network. Finally, a feature fusion module (FFM) is designed to utilize advanced semantic information, enhancing low-level detail features and improving the ability to capture subtle features. Finally, ablation and comparative experiments are conducted on a self-made bird target dataset in the background of substations. Experimental results demonstrate that this method has good performance in bird target detection scenarios in substations.

In order to accurately analyse the construction process of circular shaft diaphragm wall in geotechnical engineering, based on the advancement of geotechnical engineering technology and the finite element method, we constructed a three-dimensional finite element analysis model, which comprehensively considered the factors such as groundwater seepage, soft and hard interlocking stratum conditions. The model comprehensively analysed the influence of soil layer, soil properties and support system on the project, and deeply studied the stress-strain changes of the support structure and soil body through numerical simulation. The construction complexity was accurately simulated by reasonably setting the boundary conditions, contact conditions and soil ontological relationship model. Comparison with the actual monitoring results verifies the accuracy of the model and error analysis is carried out. The study shows that the 3D finite element model provides a scientific basis for engineering design and construction.

The few-shot intestinal polyp image semantic segmentation aims to segment unseen targets in the query image with only a few pixel-by-pixel annotated support images. However, existing few-shot intestinal polyp image semantic segmentation methods mainly mine valuable guidance information from the support branch, ignoring the role played by the task target information in the query branch in improving model performance. In this paper, a few-shot intestinal polyp image semantic segmentation method based on multi-scale cross fusion attention is proposed. Firstly, the pretrained convolutional neural network is used to map the images on both branches into the same feature space, and the multi-scale information on different branches is extracted respectively. Then, cross attention is used to establish the scale fusion between the multi-scale information of the support branch and the query branch, promoting the semantic alignment of features between branches. Finally, the similarity values between the encoding features at each position on the prototype set and the query image are calculated using a parameter-free metric method, and the unseen target area in the query image is segmented according to the similarity value. Extensive experiments on open-source intestinal polyp image dataset demonstrate the superiority of the designed method.

The assembly quality and assembly success rate of ships are crucial for meeting design specifications and operational performance. To improve shipbuilding efficiency and reduce construction costs, a Monte Carlo-based optimization design method for assembly tolerance of ships was proposed in the study, and then a related verification was carried out via a case. The proposed optimization design method was applied to the tolerance analysis of the assembly of marine pump equipment and its base. The analysis results were compared with the calculation results obtained through the extremum method and probability method. The results showed that the tolerance optimization design method allows for more lenient tolerance values for each component link compared to extremum method and probability method. Additionally, the assembly success rate could reach up to 99.89 % by adjusting the basic dimensions of component link A6 from the perspective of contribution rate. The optimization design method for the overall assembly tolerance of ships significantly improved the assembly success rate, highlighting its considerable potential for broader application.

Power equipment public opinion events can affect all aspects of the power industry, including equipment failures, power supply interruptions, environmental impacts, and policy changes. Through public opinion event extraction, electric power companies and related government agencies can monitor and analyze public concerns and feedbacks in real time, quickly respond to potential problems, improve equipment management and maintenance, increase power supply reliability, and reduce operational risks. In this paper, we propose a deep reinforcement learning-based public opinion event extraction framework for electric power equipment, and design a reward function based on the recognition of trigger words and the detection results of related event elements, and finally, through comparative experiments, we can see that the extraction results of the proposed framework are better, and it can meet the requirements of public opinion event extraction for electric power equipment.

The smart meter requires a long lifetime and high reliability. Accelerated Life Test (ALT) is an important reliability acceptance method. The switch is an important human-machine interface for electricity meters, and its reliability will directly affect the reliability of the meter. In the article, scanning electron microscopy (SEM), X-ray energy dispersive spectroscopy (EDS), and ion chromatography were used to analyze the failure mechanism and source of failure factors of the light touch switch that failed after ALT testing with the intelligent meter. The results indicate that the deionized water used in the production process of the switch is contaminated with K+ , resulting in excessive residual K+ inside the switch. In the ALT environment, water vapor enters the interior of the switch, forming a pathway between the high-voltage and low-voltage terminals inside the switch. Under the action of the electric field, K+ gradually accumulates to the lowvoltage terminals, forming potassium salts with other substances inside the switch. During the ALT cooling and dehumidification phase, the solubility of potassium salt decreases, the solvent decreases, and potassium salt precipitates. Due to the poor conductivity of potassium salts, the switch fails. The article proposes improvement methods for such failures from the perspectives of pollution source control and improving product protection capabilities, which can effectively improve the reliability of smart meters.

This study introduces a novel coarse-to-fine framework combined with ranking regression designed to capture the ordinal nature of age progression. Our approach initially categorizes ages into broader groups, utilizing the inherent order of age labels to refine age estimation hierarchically. A ranking regression model then meticulously fine-tunes the predictions, resulting in a more accurate age estimate. We present a multi-stage neural network architecture that first differentiates between broad age categories and then hones in on more specific age distinctions. Our evaluation of multiple benchmark datasets indicates a substantial reduction in prediction error over current leading models. The empirical findings highlight the effectiveness of our methodology in addressing the complex, non-linear patterns of facial aging. The proposed method propels the domain of age estimation forward and provides a versatile framework for other ordinal regression tasks.

Ensuring the accurate prediction of water supply network pressure is crucial for the efficient operation of the water supply system. The purpose is to address the issue of insufficient accuracy in short-term forecasting. The method of short-term prediction for network pressure nodes using the Particle Swarm Optimization-Back Propagation (PSO-BP) neural network algorithm is proposed in this paper. Firstly, the real pressure data collected from a water supply network, upon which this paper relies, is cleaned using the Cook’s distance method. After the dirty data is removed, Back Propagation (BP), Genetic Algorithm-Back Propagation (GA-BP), and PSO-BP are employed to predict the pressure in the water supply network. By comparing the prediction results, PSO-BP was found to be the most accurate among the three algorithms, with an RMSE (Root Mean Square Error) of 1.9132. In order to solve the problem of insensitive regions in pressure prediction, a variable sliding window method is proposed to determine the data set based on the previous method. The results indicate that this method can effectively improve the accuracy of prediction in insensitive areas.

This paper proposes an image acquisition module based on Field Programmable Gate Array (FPGA), aiming to realize image acquisition and processing. This module design mainly uses ADV7180 chip, camera OV5640, and synchronous dynamic random access memory (SDRAM) memory. This module includes functional modules such as image sensor interface, data cache, image processing, and image storage, and uses the parallel computing power of FPGA to implement image acquisition processing tasks. FPGA-based modules have significant advantages in image acquisition and processing through their parallel computing capabilities and reconfigurability. Then the design and implementation of the FPGA-based image acquisition module offers higher processing speed and lower latency than traditional software implementations

From the 21st century, the progress of science and technology becomes more and more rapid, making the application of various macro and micro sensors such as radar, infrared, photoelectric, satellite, TV camera, electron microscopy imaging, CT imaging more and more widely, and the amount, scale, and complexity of spatial data are rapidly increasing, which has far exceeded the human ability to interpret. Because end-users are unable to analyze all the data in detail and extract the spatial knowledge of interest, the phenomenon of “spatial data explosion but lack of knowledge” occurs. Therefore, in order to improve the utilization efficiency of spatial data, it is necessary to study the complex temporal data model and extraction technology of multi-spatial database and multi-spatial database. Using spatial data mining and knowledge discovery to automatically or semi-automatically mine previously unknown but potentially useful spatial patterns from multi-spatial databases becomes necessary. In view of the above situation, this paper used clustering algorithm to verify and analyze the complex temporal data model and extraction technology of multi-spatial database. By comparing the four aspects of precision P, recall rate R, comprehensive performance F, and content extraction speed of different algorithms, the relationship between the two was obtained. The experimental research results have shown that under other conditions being the same, the precision P, recall rate R, and comprehensive performance F of the k-means clustering algorithm are basically above 97%, higher than those of the ParEx and vu algorithms. In terms of the time required to extract the content, the time of the k-means clustering algorithm is 0.35s, which is much lower than the 0.49s of the ParEx algorithm and 0.61s of the vu algorithm. It can be proved that the k-means clustering algorithm can promote the development of complex temporal data model and extraction technology of multi-spatial database, indicating the positive relationship between the two.

To improve the accuracy and learning performance of lightning prediction models, a BP-ANN binomial classifier for lightning prediction based on incremental learning and spatiotemporal characteristics is proposed. By using incremental methods and learning historical data based on the spatiotemporal characteristics of the data, various BP-ANN models are established to predict and classify new data, and then the category of the new data is determined by majority voting. Three lightning prediction models were constructed: incremental learning-based BP-ANN model, spatiotemporal characteristic based BP-ANN model, and BP-ANN model combining incremental learning and spatiotemporal characteristic. The prediction accuracy and learning performance were tested on a real lightning dataset, and the results show the advantages and disadvantages of incremental learning, spatiotemporal characteristic, and their combination.

Erasure coding technology can reduce costs, improve data reliability, and enhance the usefulness of the system. This technology has been widely designed and used in distributed storage systems. However, in the case of multi-tenancy, none of the popular distributed storage systems support user-level flexible and configurable encoding modules. To solve the problem that the cost and reliability cannot be controlled by custom encoding parameters, the proof of equivalence principle under different code construction schemes is provided, and then the erasure encoding module under multi-tenancy is designed. This module allows a flexible selection of stripe ratio parameters and code algorithms, and data conversion under different algorithms. Based on the generator polynomial, the code library is provided which is optimized by avx2 instruction sets. Compared with the open-source generator matrix-based code solution, the library can resist silent errors. Finally, the analysis and evaluation of the throughput performance of the three code libraries that support avx2 instruction sets is provided. The experiment results show that ISA-L library could maintain high encoding throughout with the growth of the size of the stripe. The design of the flexible erasure code module under multi-tenancy environment has filled the gap of the distributed storage system in the refined control of cost and reliability.

In the rapid development of the Internet era, various intelligent devices will generate massive image data. With the emergence of large-scale manually labeled datasets, deep learning technology has made great breakthroughs in the field of computer vision such as image classification, image super resolution and object detection. However, manually labeling image data is a tedious and time-consuming process. In contrast, unlabeled datasets are cheap and easy to obtain from the internet. Therefore, how to effectively use the massive unlabeled data on the internet is one of the research hotspots in the field of computer vision. Self-supervised representation learning constructs supervision signals by designing self-supervised pretext tasks, and learns rich semantic representations from unlabeled datasets. However, many existing self-supervised representation learning models usually need a large batch size to learn a good visual representation in the training process, and setting a large batch size often requires a large amount of computing resources. In order to solve the problems in the above self supervised representation learning algorithm, we apply self-supervised representation learning to object detection task and propose a self-supervised object detection method based on spatial scale learning and category prediction. Without the need to add additional manual labels, we help the model to learn the spatial scale relationship and category relationship between objects in the image by introducing the spatial scale information learning task and category prediction task. Moreover, in the feature extraction stage, we use the feature pyramid network and attention mechanism fusion to help the model better adapt to the size difference of different objects in the image, so as to learn more abundant detail information in the image and further improve the performance of the model. The experimental results show that our method can achieve better performance.

Signal reconstruction from its power spectrum is an important technique widely useful in different application fields. However, uniqueness is generally hard to be guaranteed by usual approaches due to nonlinearity of the power spectrum measurement models. In this paper a very general structure of the 1-dimensonal signal with given power spectrum is established and some conditions are introduced to guarantee the uniqueness of solution in signal reconstruction on basis of this formulism. Such conditions are about one or more additional information of the signal component and two algorithms are established for signal construction. One of the algorithms is exact and another one is iterative and approximate, both can reach the real signal under the introduced conditions.

This paper presents an automatic recognition method for abnormal behaviors in exercising human bodies, including sample image collection and preprocessing. The displacement of shoulder, elbow, wrist, and knee joints of the exercising human body are taken as state variables. A normal exercising human body form observer is designed based on Kalman prediction, along with the design of observation gain for the exercising human body form observer. A threshold for determining abnormal behaviors in exercising human bodies is set. Another set of exercising human body samples, Sample II, is used to test the designed exercising human body form observer. The main advantage of this method lies in its ability to establish a database through the collection of human movements, to predict subsequent movements, and to preemptively determine the trend of any abnormal behaviors so that intervention control can be applied in advance if necessary. Additionally, the robustness of recognition is enhanced through image enhancement, image filtering noise reduction, morphological analysis of images, and edge detection processes in data.

Text-to-SQL tasks aim to bridge the gap between natural language questions and SQL queries, enabling efficient interaction with databases without the need for expertise in SQL coding. In this paper, we introduce FATO-SQL, a novel Large Language Model (LLM)-based framework designed for medium-scale LLMs to generate complex SQL queries. FATO-SQL leverages the Retrieval-Augmented Generation (RAG), prompting engineering, and two rounds of LLM calls for SQL generation and diverse response generation. We implemented FATO-SQL using production data from the petrochemical industry, testing it with multiple tables joins and multi-level nested SQL queries. Results show that FATO-SQL achieves an overall accuracy of 94% on 40 testing questions. The FATO-SQL framework demonstrates promising potential for practical industrial applications, highlighting its efficacy and adaptability in real-world scenarios.

Signal reconstruction from its power spectrum is an important technique widely useful in different application fields. However, uniqueness is generally hard to be guaranteed by usual approaches due to nonlinearity of the power spectrum measurement models. In this paper a very general structure of the 1-dimensonal signal with given power spectrum is established and on basis of this formulism some conditions are introduced to guarantee the uniqueness of solution in signal reconstruction. Such conditions are about the information of the interference signals and according to which an algorithm is established for signal construction. The algorithm is simple, efficient and its solution is unique in case of linear-phase signals. Some generalized and more practical conditions are also discussed.

This paper proposes a gene association analysis algorithm that effectively identifies causal relationships between genes through gene association entropy, and uses heuristic search strategies to construct gene association Bayesian tree (GABT) and gene association Bayesian forest (GABF). Unlike ordinary gene Bayesian networks that describe the dependency relationship between gene expression levels, GABT and GABF are a type of gene sequence Bayesian network. The object of gene association analysis is the sequence formed by sorting the gene expression values of biological tissue samples and replacing them with gene column subscripts. The experimental results on multiple tumor or non tumor gene expression datasets show that the Bayesian network classification algorithm based on gene association analysis can better fit gene expression data than other similar algorithms, with significantly improved accuracy or reduced analysis time

This study employs the LISA-SDE Hybrid Modeling Approach to analyze the distribution and dynamics of cold chain logistics enterprises across China from 2018 to 2022, focusing on enterprise numbers, storage capacities, and vehicle fleets. Results from the LISA analysis indicates strong, persistent spatial clustering in regions such as Gansu and Qinghai, with Gansu maintaining high LISA indices from 0.798 in 2018 to 0.789 in 2022 and consistently low p-values below 0.005. The Standard Deviation Ellipse analysis reveals a significant geographic center shift northward by approximately 21 kilometers between 2018 and 2019 and an increase in the spatial extent of logistics facilities, as evidenced by the expansion of the ellipse area from 228.417 square units in 2018 to 231.006 square units in 2021. These findings highlight the dynamic evolution of China’s cold chain logistics infrastructure, underscoring significant advancements in spatial distribution and strategic clustering, which are crucial for informing infrastructure development and policy-making aimed at modernizing the system by 2035.

E-sport project League of Legends is a top event with a complex mechanism, in order to improve accuracy win rate prediction of League of Legends, the logarithmic energy entropy generates expansion vectors as grouping strategy to enhance contextual information between use of heroes with stronger relationship features, using regulating force mechanism to assign heroes feature weights, extracts heroes lineup selection and role relationship position signal features in turn, establishes a win rate prediction model to combine convolution neural network as a classifier to fully extract the deeper features. Simulation results show that the proposed grouping strategy and feature combination strategy can obtain an overall prediction accuracy 75.32%, which can effectively improve the prediction model accuracy.

With the continuous development of 5G technology and the increasing number of application fields, the requirements for network stability and speed have also been raised. Once the network is affected, it will not only affect normal work and life, but in severe cases, it can also cause major problems such as information omission and false alarm. This article first analyzes the development background of 5G networks and points out that the first to fifth generation mobile communication systems have successively implemented functions such as calling, text messaging, internet access, video, and intelligence. 5G, due to its advantages of large-scale IoT connectivity, critical task applications, and enhanced mobile experience, has a significant impact on industries such as healthcare, manufacturing, transportation, and entertainment. Secondly, through the calculation of peak rates for 5G uplink and downlink, five main factors affecting 5G rates are identified, namely time slot modulation degree GRANT, RB scheduling number, MCS order, RANK, and BLER. Finally, through the analysis of low RB scheduling and rate non-compliance examples, it is demonstrated that 5G rate calculation can be effectively applied to the analysis of network influencing factors.

In the database system based on hybrid storage, the hot and cold data identification algorithm aims to distinguish the hot and cold data when data migration occurs and eliminate the cold data to the storage media with lower performance, and its accuracy directly affects the access efficiency of the hybrid storage system. The traditional hot and cold data identification algorithm mainly identifies hot and cold data based on single features such as access time and access frequency, which has certain limitations. For this reason, a hot and cold data identification algorithm for database based on heat is proposed, which integrates two important features of data access frequency and recent access time to quantify the heat of data, and distinguishes hot and cold data based on this heat. The experimental results show that the algorithm has good stability under different conditions and can improve the cache hit rate by up to about 8% compared with the traditional cache replacement algorithm.

Target track quality grading system plays an important and practical role in real-time evaluation of target trajectory quality and improvement of the target tracking accuracy of detection equipment. In the paper, the definition and distribution of target trajectory accuracy is introduced, and the multi-dimension trajectory state error is uniformly quantified using error sphere radius. Then the constructing process of target track quality grading system is designed. At last, simulations show the affection of grading thresholds and target trajectory accuracy calculation models on the construction of target track quality grading system.

In the construction of clean and low-carbon community energy management systems, the challenges of the new era of low-carbon community development lie in monitoring and utilizing the massive panoramic information obtained from physical equipment within the community, constructing digital twin models of intelligent clean energy systems, and achieving multi-energy complementarity and energy conservation and emission reduction. Based on the optimization and complementarity of multi-energy interconnected systems and the coordinated operation control technology of “sourcegrid-load-storage,” this paper designs a digital twin-based green and low-carbon community energy system operation management platform that integrates the needs of multiple business scenarios. By integrating modeling, simulation, artificial intelligence, and hierarchical control across multiple temporal and spatial scales through digital twin technology, this platform offers functions such as energy planning, operation and control, demand response, distributed energy trading, and carbon neutrality services. It addresses the issues of information barriers hindering the development of clean energy systems such as “source-grid-load-storage integration” in energy systems, effectively supporting key aspects of integrated energy systems, including planning and design, optimized operation, online safety analysis, and equipment health management.

Gesture-based interaction represents one of the most intuitive and immediate methods for human communication with their surroundings. The role of gesture recognition technology is particularly significant in fields like sign language interpretation and human-computer interface. Flexible sensors, characterized by their high sensitivity, excellent stretchability, and affordability, are particularly well-suited for incorporation into wearable devices. This paper provides a comprehensive overview of gesture recognition systems developed in recent years that utilize flexible sensors. It delves into the hardware architecture of notable data gloves and emerging wrist-worn devices. The study employs traditional machine learning techniques for recognizing both static and dynamic gestures. Furthermore, it explores the creation of advanced deep learning models, utilizing Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) architectures. Additionally, the paper examines the potential future applications of gesture recognition systems, highlighting their utility in smart home environments and surgical training. In conclusion, the article identifies existing challenges in terms of environmental robustness of sensors, signal latency, and the need for enhanced data set development.

Water supply forecasting methods can be categorized into two primary types: traditional forecasting methods and machine learning methods. Traditional methods have limited forecasting accuracy for daily water supply, while machine learning methods have better model description ability and can find data details that are difficult to capture by traditional algorithms. In this paper, a water supply forecasting method based on the Long Short-Term Memory (LSTM) network is proposed. The water supply data of Xinmin Water Plant in Dianjiang County from March 1, 2024, to March 30, 2024, are used to verify the method. The error analysis of the error results shows that the RMSE is 0.051718, and the NSE is 0.960363. The forecasting of rural water supply based on LSTM has high forecasting accuracy and stability, which is an effective forecasting method.

Battery energy storage technology is crucial for advancing new energy solutions, achieving carbon neutrality, and promoting sustainable development. Effective monitoring and management systems are essential components of this technology. This paper presents the design and implementation of a new energy storage information universal management system. The proposed system offers real-time monitoring of battery voltage and temperature via an intuitive visual human-machine interface, while also enabling authority management. It further supports remote protection and control operations, including battery access and connection management. Moreover, the system is capable of systematically collecting and analyzing battery performance data on a regular basis.support IEC 61850/IEC 104 background monitoring, and enable power conversion with PCS. Experiments on multiple energy storage power stations have shown that the system significantly improves the management efficiency and safety of battery energy storage, improves user interaction, and reduces maintenance costs.

An improved Pan-Thompkins R-wave localization algorithm is proposed to solve the problem of performance degradation when Pan-Thompkins algorithm is used to locate R-waves in electrocardiogram (ECG) signals of arrhythmia. The algorithm introduces new rules in the threshold adjustment stage to better capture the trend and pattern of the signal; the reverse search mechanism is added in the predefined interval, and the R peak that may be lost is detected by the new rule. In addition, the RR interval check mechanism is added to prevent the T wave from being mistakenly detected as the R wave in the case of complex signal modes. The MIT-BIH arrhythmia dataset and the MIT-BIH noise pressure test dataset are used to test the proposed algorithm with the Pan-Thompkins algorithm and three other commonly used R-wave localization algorithms. The results show that the proposed algorithm has high accuracy and also shows good robustness in the case of arrhythmia.

The fundamental, supportive, and leading roles of standards in industrial development is becoming increasingly prominent, making them a key tool for driving industrial innovation and development. This paper utilizes machine splitting technology and positioning assignment technology to process and analyze standard documents from Haidian District, Beijing. By calling relevant interfaces, the contribution of different types of drafting units in Haidian District to standard formulation in various industrial fields is tested and analyzed, revealing the characteristics of standardization work among different entities in Haidian district.

Circular shaft diaphragm walls are mostly circular, but the existing theories are mostly based on rectangular or planar structures, which lack support for the design and construction monitoring of circular shafts. In this paper, based on actual cases, the parameters such as water-soil pressure, wall strain and soil displacement behind the wall during the construction of circular shaft diaphragm walls are analyzed to explore their mechanical behavior and force characteristics. The results show that the circular vertical shaft diaphragm wall has unique mechanical behavior and force characteristics, which are different from the traditional theory. The earth pressure is maximum at the depth of 4 m, which is far more than the value of Rankine’s theory; the ground stacking load has a significant effect on the earth pressure; the wall strain and the soil displacement behind the wall are small, which shows a good self-stabilizing effect. The conclusions of this paper are of great significance to optimize the construction and improve the safety, and provide reference for similar projects.

With the rapid development of personalized medicine and precision treatment, tumor single-cell sequencing technology has highlighted its important role in the study of tumor heterogeneity. This article systematically reviews the technological progress and current applications of single-cell sequencing technology in the field of tumors, and explores in depth how artificial intelligence (AI) can promote feature optimization of single-cell data and its potential value in tumor diagnosis and treatment. This study introduces the current mainstream data optimization methods and analyzes how artificial intelligence technology can improve diagnosis and treatment strategies through case studies, pointing out the advantages of artificial intelligence in improving the accuracy of single-cell sequencing, especially in enhancing tumor biomarker identification, predicting recurrence potential, and assisting personalized treatment decision-making. This article concludes that artificial intelligence technology has broad prospects in optimizing precise tumor diagnosis and treatment, effectively promoting the transformation of tumor management, and providing new ideas for the innovation of future tumor treatment models.

With the widespread application of 5G technology, user demand is constantly increasing, and the user base is also growing. How to allocate core network resource pools reasonably based on user size and demand, and effectively utilize resource efficiency has become an urgent problem to be solved. This article first analyzes the development history of the core network and points out that the evolution of the core network has gone through stages of 2G, 2.5G, 3G, 4G, and 5G, gradually achieving an increase in network speed. It is also pointed out that the scale of 5G base stations and 5G users in the first half of 2024 will increase by varying degrees compared to the same period last year. Secondly, the factors affecting the 5G core network resource pool were analyzed, including technological development, business requirements, cost considerations, architectural flexibility, maturity, and reliability. According to the calculation formula, provide a method for calculating network bandwidth and configuring resources such as port firewalls. Finally, the network bandwidth and port configuration of three different provinces were calculated through examples.

Chinese characters, as the fundamental medium of communication within Chinese culture, stand out due to their intricate structures. Strokes, the basic elements of Chinese characters, are crucial for assessing Chinese handwriting. Accurate stroke extraction is essential and serves as the initial step in this evaluation. Traditional stroke extraction methods typically rely on specific rules that often fail to capture the full complexity of Chinese characters and cannot align the strokes according to the sequence used in template characters during assessments. To address these challenges, this paper redefines stroke extraction as a multi-label semantic segmentation task and introduces a new model, M-TransUnet. This model utilizes a deep convolutional approach to train individual Chinese characters, maintaining the integrity of stroke structures and resolving ambiguities in stroke segment combinations. It also accurately determines the order of strokes, aiding in subsequent tasks such as stroke evaluation. Furthermore, since handwriting images are only segmented into foreground and background without additional color cues, they are prone to false positive (FP) segmentation noise. To mitigate this issue, we propose a Local Smooth Strategy on Strokes (LSSS) that diminishes noise impacts on the segmentation results.

The research, development and application of metal additive manufacturing machine tools are important development directions for future intelligent manufacturing. However, as an important component of metal printer machines, the workbench still has problems that need to be solved, such as excessive weight, insufficient static and dynamic performance, etc. This paper takes the 3D SVT machine tool as the research object to simulate the actual working conditions of the machine tool for static analysis and modal analysis, and studies the relationship between the static stress deformation of the workbench and the natural frequency value; and conducts simulation analysis with the goal of determining the minimum mass and maximum frequency of the workbench, and performs variable density method topology optimization design; to determine the optimal solution of removing the structure, adjusting the rib plate size data and adding fan-shaped rib plates. This method effectively lightweights the machine tool workbench and improves the natural frequency value, which can effectively improve the working efficiency of the 3D SVT machine tool and promote the research and development and application of metal additive manufacturing machine tools.

Handwritten digit recognition is a typical application of computer vision, and its results can be widely used in the fields of zip code recognition, statistical report recognition, and test score determination. Handwritten digit recognition is still a hotspot in image recognition and classification, and the deep learning algorithm based on convolutional neural network (CNN) has the structural characteristics of local region connection, weight sharing, and down sampling, which makes convolutional neural network have an excellent performance in the field of image processing. In the paper, the adaptive binarization method is used to realize the segmentation of handwritten digits and background, the individual digits are segmented and extracted sequentially using the improved algorithm based on directional projection, the LeNet-5 model of convolutional neural network is trained by the handwritten Minist training dataset, and the segmentation and recognition of multiple handwritten digits within a single image is realized using TensorFlow. The experimental results show that the method in the paper has high reliability, and the average recognition rate of the trained model for new handwritten digits is above 92%, which achieves the expected results.

This research paper provides a pragmatic basis to use quaternions in application to machine learning and assesses cat and dog image classification with a twist of using OBLR (Orthogonal Basis and Logistic Regression) with PCA (Principal Component Analysis) and Entropy. The OBLR method uses the Gram-Schmidt technique of orthogonalization to derive an independent, orthogonal feature space for cats and dogs separately, followed by application with logistic regression for categorization. In simple terms, “The Entropy and PCA Method” reduces the data dimensionality in the image with PCA and classifies based on the variation of entropy among classes. It picks the limitations of these two methods and chooses the advantages. The result of the study indicated that both the methods considered were found effective in the classification of images containing cats and dogs, and therefore, useful applications in the management of stray animals and control of pet health. Our forthcoming scopes will include the implementation of OBLR and PCA with entropy, the implementation of more sophisticated machine learning models like CNNs, and statistical metrics of feature selection techniques that would be useful to further improve the classification accuracy.

In this paper, we establish multi-objective optimization model and water level prediction model through multi-task learning algorithm, entropy weight analytic hierarchy process (AHP), convolutional neural network (CNN) and Long Short-Term Memory (LSTM) neural network, etc., and calculated weights of 0.665, 0.761, 0.866, and 0.122 for water level elevation, river flow, cyclic variation, and water level stabilization, respectively, and Lake Ontario’s optimal water level of 74.926 m.

When deciding on the path of a truck carrying hazardous chemicals, it is imperative to consider not only the associated transportation costs but also the imperative to reduce the risk. Grid network is a typical urban road network structure. It is necessary to consider the flexibility afforded by parallel roads when conducting harmful chemicals transportation. This paper studies the path selection of harmful chemicals transportation on a grid network. The grid network is formally defined, and a path selection problem is introduced along with its mathematical representation. A fast solution method is proposed based on the characteristics of the grid network. Taking the local network in Xi’an City as an example, the effectiveness of the proposed solution method is verified.

The technology of generative general artificial intelligence not only revolutionizes machine intelligence but also plays a significant role in enterprise digitalization. Given that natural gas sales companies offer conventional human-centered services, cost and efficiency have emerged as a set of irreconcilable contradictions. The automated intelligent customer service model, based on the new generation of artificial intelligence technology, is like an open key to high-quality development. We propose a method to create a knowledge graph of customer service business processes by guiding a multi-modal large language model to generate the knowledge graph with prompt words. Based on the open-source multimodal large language model, the general ability of the large model was applied to identify, analyze, and extract the business process table in the “Natural Gas Customer Standardized Service Business Process Guidebook” through customized task prompt design. Ultimately, we successfully constructed a business process knowledge graph, confirming the feasibility and effectiveness of this method. This method aims to optimize intelligent customer service by providing high-quality answers. The method enhances automation, intelligence, and standardization of customer service, which improves work efficiency and controls costs simultaneously.

In order to overcome defects resulted from replacing the photo background color, an improved method is proposed for background replacement. The α-values in the alpha matte are transformed to enhance the details in the transition area. The new photo is composited by comprehensively considering the foreground portrait, new background, old background and the alpha matte. The experimental results show that the foreground portrait appears clearer and more complete, and the transition area between the foreground and background is more harmonious and more natural.

In the field of geophysical exploration, writing seismic acquisition design reports is an important yet tedious task. How to improve the efficiency while ensuring the quality has become a significant challenge for seismic acquisition designers. This paper proposes a seismic acquisition design report generation technology solution based on generative artificial intelligence. First, it divides documents into multi-level headings to create a repository of historical design reports and supports convenient report retrieval functions. Then, leveraging the powerful learning and reasoning abilities of the large language model, in-depth analysis of historical design reports is conducted, and new technical reports are automatically generated based on this. This innovative solution greatly improves the work efficiency of seismic acquisition designers. The information collection and report writing work, which originally took more than a week to complete, can now generate a high-quality report in just a few minutes. This technical solution can not only be used for seismic acquisition design but also provides effective exploration and practice for the digital transformation of related industries. In the future, with the continuous development and improvement of technology, this technological solution will demonstrate greater application potential and value in various fields.