This PDF file contains the front matter associated with SPIE Proceedings Volume 13259, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

In recent years, with the implementation of sustainable development policies and the rapid advancement of science and technology in China, new energy vehicles have garnered widespread attention, leading to a continuous expansion of the market size. At present, there could be shortcomings in the motor and battery control technology of new energy vehicles when being used, which has a significant impact on the effectiveness of motor drive control and is harmful to the safe and stable operation of the vehicle. This paper delves into discussing various types of faults in hybrid control technology for new energy vehicles and proposes fault handling strategies for motor drive control technology in order to facilitate effective motor drive control for new energy vehicles.

This paper utilizes the case study of instability in the tracking operation of a specific high-power servo system to develop an improved fault diagnosis methodology. By employing the fault tree analysis method, hierarchical weights are determined using exponential scaling as the scaling method within the analytic hierarchy process. Subsequently, the weights for the causes of poor tracking performance in the high-power servo system are identified. Finally, fuzzy fault diagnosis methods are employed to enhance the precision of these fault cause weights, resulting in more rational and applicable weights for diagnosing various sudden faults during the operation of high-power servo systems.

In response to the existing problems in the fault diagnosis technology of ship electric propulsion systems, based on fault tree analysis, this paper proposes an ant colony algorithm based fault diagnosis method for electric propulsion systems, which overcomes the problems of the diagnostic system's inability to learn, low diagnostic efficiency, and poor accuracy in the current fault diagnosis process. On the basis of in-depth research on ant colony algorithm, fault tree analysis method, and ship electric propulsion system faults, this method combines the two algorithms and introduces user feedback information based on the actual situation of the ship.

The conversion from the source protocol to the target protocol is achieved by mapping the header of the source protocol to the header of the target protocol. A single conversion process from the source protocol to the target protocol is referred to as a session. In a heterogeneous protocol conversion system, there are complex mapping relationships between different source and target protocols. To address the complexities of mapping relationships between heterogeneous protocols and the difficulties in session management, a session management system for heterogeneous protocol conversion was designed and implemented. Experiments show that the designed session management system can establish a protocol mapping rule table and extract the protocol mapping rules corresponding to the current session. This management system achieves efficient and flexible session management, providing support for protocol conversion.

The impact of GPUs on people's lives is growing. Artificial intelligence applications are everywhere. Model inference is one of the main ways AI applications serve humanity. The current situation is that the utilization rate of GPUs in smart computing centers is low, and idle resources lead to a lot of waste. How to efficiently use GPUs to complete model inference tasks is a problem that AI practitioners commonly encounter. GPU virtualization technology can improve the utilization rate of GPUs and improve indicators such as inference throughput and latency. It can be foreseen that GPU virtualization is one of the key technologies in the development of GPUs. There are mainly two implementation methods for GPU virtualization technology: vGPU and MIG GPU. This article organizes the two key indicators of task completion time and throughput through experiments, analyzes the strengths and weaknesses of the two virtualization technologies, and their possible application scenarios. It provides a reference for AI practitioners to use GPU virtualization technology in the future.

Articulated vehicles are usually connected by a rigid rotating joint between the front and rear fuselage, and in actual working conditions, the fuselage will inevitably produce side slip. In this paper, the dynamics of articulated vehicles affected by sideslip are analyzed and modeled, and a non-holonomic vehicle kinematics model with sideslip Angle and the corresponding rotational error dynamics model are established. A prediction controller is designed according to the actual working conditions. The controller switching strategy is selected by the change of sideshow Angle, and the sliding error is compensated to provide the overall running stability and improve the tracking performance of the vehicle. Through the relationship between the actual pose and the reference trajectory, the control objective function is optimized to achieve the lowest deviation of the vehicle heading Angle, so as to obtain the vehicle speed and link Angle control quantity, so as to solve the multi-variable and multi-constraint state. Through simulation, the results show that the prediction controller can solve the sideslip problem when the vehicle frame is offset, and can compensate the error well, which proves that the scheme can deal with the sideslip problem well in the actual working condition.

Rolling bearings are one of the key components for the stable operation of many mechanical equipment. In some high-noise industrial environments, the collected bearing signals are usually affected by environmental noise, which leads to the distribution difference between labeled training data and unlabeled test data, resulting in a decline in diagnostic performance. The ViT model of Transformer encoder is proposed for rolling bearing fault diagnosis. The input time-domain vibration signal is segmented into small blocks based on a predetermined size. The self-attention mechanism is then employed to integrate global vibration information for fault diagnosis. After training with augmented data, the ViT model has high fault diagnosis accuracy in high noise environment. In addition, the t-SNE network visualization technology is used to show the final feature extraction results of the neural network, which improves the interpretability of the neural network. The input time-domain vibration signal is segmented into small blocks based on a predetermined size. The self-attention mechanism is then employed to integrate global vibration information for fault diagnosis.

Transformers are a crucial part of the power system, and accurately evaluating the insulation life of oil paper is very important for the stable operation of the power grid. By analyzing FDS spectra with different degrees of aging, multidimensional characteristic quantities that can characterize the aging condition of insulation paper were extracted. Based on the long short-term neural network, a quantitative relationship between frequency domain characteristic parameters and aggregation degree values was established. Finally, combined with the cumulative loss dynamics equation, a prediction model for the insulation life of oil paper was constructed. The results indicate that the lifespan prediction model can accurately predict the lifespan of oil paper insulation, providing data support for the daily operation and maintenance of transformers.

Aiming at the current fault characteristics of rolling bearing vibration signals with non-linear and non-smooth characteristics that are not clear, a rolling bearing fault diagnosis method based on improved support vector machine (SVM) is proposed. The Improved Whale Algorithm (IWOA) is used to optimize the support vector machine model parameters in order to create the best IWOA-SVM identifying faults model for motor bearing defect classification. According to the results of the study, the use of the IWOA-SVM classification model demonstrated significant results in fault diagnosis with an accuracy of 99.33%.

This paper presents a hybrid topology wireless charging system for electric vehicles. Firstly, the series (S-S) compensation network and double capacitor and inductor in parallel (LCC-S) compensation network are analyzed. Then, the hybrid compensation topology is proposed. Finally, simulation verification is carried out. The results show that the hybrid compensation topology of LCC-S and S-S has better efficiency and safety.

The miter gate of large ship lock is one of the core control technology requirements of the operation control system, and it is also an important guarantee for the safe and efficient operation of the ship lock. In this paper, we establish the hydraulic-electric control system model of closing alignment of miter gate of a large ship lock by using the principles and methods of hydraulic compensation and PLC control. We realize the function of adjusting and correcting the closing alignment of miter gate, and ensured the safety of lock equipment and operation.

In this study, a fault diagnosis method combining the Pelican algorithm and BP neural network is proposed for improving the fault diagnosis accuracy and efficiency of microgrid systems. Pelican algorithm is a new optimisation algorithm developed in recent years, which can effectively optimise the weights and deviations of BP neural networks. Based on this characteristic, the Pelican algorithm is used to process the data of the actual microgrid system, and compared and analysed with the traditional BP neural network and RNF neural network. The experimental results show that the BP neural network optimised by the Pelican algorithm exhibits higher accuracy and robustness in microgrid fault diagnosis, which provides an important support for the reliability of microgrid systems.

In recent years, with the rapid development of artificial intelligence and computer vision technology, vision based intelligent detection and positioning systems have been widely applied in various fields. This paper proposes an intelligent detection system based on machine vision to address the challenge of automatic recognition of salt cores and positioning of salt core holes during the casting process of gasoline pistons. The system includes functions such as salt core grabbing, salt core hole searching, and salt core crack detection. Based on the distribution characteristics of salt cores and salt core holes, two sets of camera partition detection methods are designed to simplify the design of robots and servo motion systems. The actual application results show that the positioning accuracy of the visual inspection system reaches within 0.02mm, and the repeated positioning performance is good, which can meet the application requirements of industrial sites.

Humanoid robot has broad application prospects in services, entertainment, medical care, rescue, and other fields. Therefore, research on the stable walking problem of humanoid robot is particularly important. To solve this problem, this paper proposes a gait control method for humanoid robot according to reinforcement learning. We established a linkage model for humanoid robot and simplified it into a linear inverted pendulum model which was used for motion control of humanoid robot. Using DDPG network as an intelligent agent to train humanoid robot in gait control in a simulation environment. Due to the combination of DDPG and DQN structure, the stability and convergence of strategy network and critical network have been improved. The MATLAB simulation results show that under the action of the DDPG algorithm, the humanoid robot can walk smoothly along the predetermined trajectory through reinforcement learning.

In this paper, a robust prescribed time control strategy is constructed to solve the torsional vibration problem of the rolling mill main drive system. Considering the nonlinear and uncertain factors in the rolling mill, the nonlinear torsional vibration model of the rolling mill is constructed by using the concentrated mass method. First-order filter is used to solve the parametric explosion problem. A performance function and state transition function are constructed to limit roll speed tracking error performance and convergence time. A robust prescribed time controller is designed to ensure that the rolling mill torsional vibration is controlled within the prescribed time. Finally, matlab software is used for simulation analysis, and the effectiveness of the designed control strategy is verified.

The problem of lateral vibration deterioration caused by speed increase of high-speed trains is becoming more and more serious. Semi-active suspension is an effective method to improve vehicle lateral vibration. The classical semi-active control of canopy can control the lateral vibration of train well. However, it has no obvious improvement effect on the vibration of the car body in the middle and high frequency range. Based on the analysis of the vibration transmissibility of the simplified physical model of 1/4 vehicle, this paper puts forward a new semi-active control strategy of improved skyhook. In this paper, the co-simulation model of lateral semi-active control of high-speed trains is established, and theoretical simulation research is carried out. The simulation results show that the improved semi-active control strategy put forward in this paper can better improve the lateral vibration of the car body compared with the classical semi-active control of the sky-hook. Therefore, this paper offers a new method for lateral vibration of high-speed trains.

The study addresses the backend trajectory planning problem in unmanned aerial vehicle (UAV) path planning, introducing a spatiotemporal parameterization method for UAV flight trajectories. It incorporates a multi-stage unconstrained minimum control (MINCO) effort spatiotemporal parameterization approach, models the trajectory optimization problem, designs penalty functions, and constrains modeling to transform the multi-constraint trajectory optimization problem into an unconstrained optimization problem, thereby obtaining the spatiotemporal optimal trajectory. Simulation validates that the MINCO algorithm effectively utilizes the degrees of freedom of each trajectory segment in adjusting time, thereby achieving lower conservatism compared to traditional trajectory representation methods and reaching closer to performance limits, resulting in the generation of efficient flight trajectories.

In the industrial production process, to address the issues of frequent manual adjustments, lengthy parameter tuning, and the challenge of selecting the best parameters in traditional PID control algorithms, this paper introduces an adaptive PID control method for heating furnace temperature control based on an improved genetic algorithm. By utilizing a chaotic algorithm to enhance the genetic algorithm's global optimization capabilities, this method achieves the optimal solution for the fitness values and identifies the optimal PID parameters for temperature control in the heating furnace. The results show that the improved genetic algorithm adaptive PID control has rapid response speed, strong anti-interference capabilities, and good robustness, thereby achieving better control effects. This method provides a theoretical basis for natural gas pre-treatment and has promising applications in engineering practice.

In this paper, the state-feedback robust stabilization of nonlinear time-lagged continuous systems is investigated by modeling the nonlinear time-lagged systems, applying Lyapunov stability theory and LMI toolbox, and designing the fault-tolerant controllers for nonlinear continuous systems by using the memoryless state-feedback controllers. Sufficient conditions for the system to be asymptotically stabilized in the case of actuator failure and the corresponding design method of the state feedback robust fault-tolerant controller are given. Finally, the reliability and effectiveness of the method are verified by numerical examples and simulation results.

In this paper, there are many troubles in the study of secondary collision of flexible structures, and it is not enough to rely on the support of theoretical knowledge. Equivalent experiments and data analysis and calculation are needed to further verify. Of course, the secondary collision of different flexible structures, their own structural stability, life span and other factors affect the experimental results. In order to more intuitively analyze the secondary collision phenomenon, a set of experimental equipment using rigid ball collision beam is reformed. The main research direction of this paper is the effect of the initial velocity and location of the rigid ball on the secondary collision. From the experimental data analysis, when the initial velocity of the rigid ball is different, the times and time interval of secondary collision phenomenon are also different.

With the widespread application of welding technology in the construction of large metal components, the need to address the challenges of low efficiency, limited output power, difficulty in achieving the required temperature at the welding head, and susceptibility to soldering defects in traditional welding power supplies has grown. To address these issues, this study introduces a welding power supply system based on an LC resonant converter. By leveraging resonance techniques, the system reduces switching losses, enhances power factor, and achieves zero-voltage switching for the main power devices, thereby elevating power supply efficiency and stability. Experimental results demonstrate that the proposed system achieves high-efficiency and high-power output, simultaneously attaining the desired welding head temperature and substantially minimizing instances of soldering defects.

In order to improve the rationality of the layout of the RGV track system, a comprehensive global layout optimization method for the RGV track system based on an autoregressive model was designed. Plan the overall layout route of the RGV rail system to ensure that the rail layout can adapt to different loading and unloading needs. We have constructed an autoregressive model for optimizing the overall layout of the RGV rail system to achieve higher transportation efficiency, lower energy consumption, and less congestion. Optimize the overall layout of the RGV track system through a multithreaded operation mechanism, achieving layout optimization. The calculation example shows that this method has good layout optimization effect and can be applied in practical production.

The rapid progress of modern agricultural science and technology has made greenhouse cultivation an important way to break the seasonal barriers and improve agricultural production. However, the complexity of the agricultural greenhouse management process and the need for precise control of environmental factors such as light, temperature, humidity, etc., do not match with the manual operation under the traditional production methods, resulting in the economic benefits of greenhouses being difficult to meet expectations. In this regard, this paper will design and construct an agricultural greenhouse intelligent control system based on the requirements of the actual project, utilizing advanced technologies such as sensors and the Internet of Things, with the aim of promoting the innovation and upgrading of agricultural production methods. Practice has proved that this system takes the perception layer, transmission layer and application layer of the Internet of Things as the structural framework, integrates data collection, processing, analysis and control functions into one, realizes real-time on-line monitoring of the main environmental parameters of the greenhouse, and completes automatic control of the actuator. This system provides solid technical support for the automation, intelligentization, networking and scientific production of agricultural greenhouses, and is of positive guiding significance for upgrading the modernization level of agricultural greenhouses in China.

Bearings, as key components in rotating machinery, can experience failures that severely impact the normal operation of the equipment and the safety of production. In light of this, the paper proposes and implements a Python-based system design for bearing fault diagnosis and prediction. Initially, the bearing data is collected using displacement sensors and acceleration sensors. Then, in combination with FFT, wavelet transform, and other methods, the collected vibration signals undergo noise reduction processing to achieve the desired processing effect. Features reflecting the state of the bearing are extracted from the time, frequency, and time-frequency domains. Subsequently, machine learning algorithms such as Support Vector Machine (SVM), Random Forest (RF), and Neural Network (NN) are chosen for training and validating the fault diagnosis model. A comparison of multiple models is conducted to evaluate the performance of various algorithms. In the realm of fault type prediction, the prediction model is built on an Adam-optimized neural network. The prediction outcome enables the assessment of the deterioration trend of non-faulty bearings, thereby aiming to preemptively address potential hidden risks associated with bearings.

This article involves the field of warehouse control technology and introduces an automatic loading and unloading warehouse control system and its loading and unloading equipment. The system is equipped with a parking car information acquisition and processing module, a cargo transportation intelligent control module, a warehouse management information comprehensive judgment module, and a data control database. Firstly, it performs parking car information acquisition and preprocessing to determine whether the loading and unloading AGV can be enabled. Secondly, it screens the first channel for car cargo transportation, and comprehensively calculates the abnormal values of warehouse management information control on the intelligent warehouse control platform. It also provides autonomous warning control prompts for warehouse management information, which is beneficial for achieving rapid loading and unloading operations of goods, greatly improving the efficiency of intelligent warehousing. At the same time, it can not only reduce the labor cost of warehousing operations, but also improve the quality and accuracy of cargo handling.

Real time acquisition of vehicle attitude information and contour shape in space is crucial for precise monitoring and control of vehicle status, which is crucial for automated loading and unloading. This article trains, optimizes, and verifies the preset detection model by constructing a parking detection dataset. Then, the parking data of the closed-loop vehicle to be tested is obtained through the parking detection model, and the parking accuracy index is calculated. Finally, based on the parking accuracy index, the parking behavior of the closed-loop vehicle to be tested is adjusted to meet the preset parking requirements, improving the accuracy of vehicle attitude contour monitoring and control, and solving the problem of low accuracy in vehicle attitude contour monitoring and control in existing technologies.

In large angular motion or other dynamic environments, the size effect can cause additional velocity error, which will consequently reduce navigation accuracy. In this paper, a calibration method for size effect error is discussed, and navigation experiments of swing motion under static base condition are conducted, which reveal that in a large angle motion environment, after compensating for inner lever arm (ILA) parameters, both velocity error and positioning error have a significantly decrease, with an improvement in positioning accuracy of over 30%.

This paper aims to address the challenges in power grid fault diagnosis by proposing a fault diagnosis model based on federated learning, called the Federated Conditional Strategy Model (FedCP). Fault diagnosis is crucial for maintaining the stability and security of power grid operations. However, the distributed nature and privacy concerns of grid data present significant challenges to the design of diagnostic algorithms. To tackle this issue, we utilized the advantages of federated learning to design a distributed fault diagnosis model that allows multiple grid regions to participate in model training without sharing sensitive data. Experimental results show that the FedCP model outperforms traditional centralized models, such as FedPer and FedRep, in terms of accuracy and F1 scores. Furthermore, the FedCP model also provides robust data privacy protection and generalization capabilities, offering a reliable and efficient solution for power grid fault diagnosis. The results of this study are significant for enhancing the safety and reliability of power grid operations and offer valuable insights for the application of federated learning in other fields.

This paper proposes an event-triggered terminal sliding mode controller (ET-TSMC) for brain-controlled wheeled mobile robot (WMR), which improves the control performance and resource efficiency. First, the ET-TSMC method is formulated by taking into account the necessary conditions for reaching the sliding surface, and an event generator is designed with the objective of conserving resource utilization. Second, we analyze the practical sliding mode dynamics and develop matrix inequality conditions to guarantee the stability of the state. The human-in-the-loop (HIL) simulation results show that the proposed controller can improve the control performance of the brain-controlled simulated WMR under external disturbance while efficiently conserving computational resources.

This paper investigates the finite-time group consensus problem of multi-agent systems with leader-following control. Finite time group consensus means that all followers in each group can be tracked separately to their corresponding leaders for a finite time. A control scheme is designed to solve the problem of finite-time group consensus. It is found through rigorous proofs that the control scheme can address consensus problem for both linear and nonlinear multi-agent systems. Finally, simulation examples verify that the state trajectory of each group of agents can converge with its leader respectively in finite time.

The frequency range of seismic waves typically extends from 0.01Hz to 100Hz, and in most cases, the signals are faint, making them susceptible to noise interference. To minimize noise interference and extract faint seismic signals, this paper proposes a method utilizing a second-order oscillating structure seismometer, employing differential capacitance detection and dual-carrier modulation-demodulation to measure tiny displacements of inertial components. In response to the shortcomings of the open-loop structure of the seismometer and the system limitations inherent in capacitance detection principles, a Linear Quadratic Gaussian (LQG) controller is designed to achieve feedback control of the system. Comparative experiments with open-loop systems, as well as those employing PID and LQR controllers, demonstrate that LQG not only improves the dynamic performance of the system but also effectively suppresses external noise. This research provides insights into the design of seismometer detection circuits and feedback controllers.

The robust control problem of permanent magnet synchronous motor (PMSM) system with uncertain parameter is studied. First, the state space model of the PMSM system is builded. It is verified that in the certain conditions, the robust control problem can be transformed into an optimal control problem to form a state feedback control. In addition, we propose an online reinforcement learning (RL) algorithm to design a controller that can adjust the motor speed to a desired level without solving the conventional algebraic Riccati equation (ARE). And the convergence of the RL algorithm is proved. Finally, the effectiveness of this control method is proved by a numerical simulation example.

The control system of the adaptive active suspension system was built, which used the hydraulic system as the balancing method, and used STM32, remote control, gyroscope, oil pump motor, solenoid valve, relay module, high-precision current analog output module and other modules. Kalman filter, PID and other algorithms are used to realize the automatic leveling of the adaptive active suspension. The automatic leveling of the adaptive active suspension is tested, and it is shown that after setting the angle deviation threshold, the control system and the applied algorithm can quickly and accurately adjust the deviation angle of the whole vehicle within the threshold. The leveling control system can be widely used in emergency rescue, agricultural picking and other fields, and has strong practical significance.

In order to ensure the stability of the operating condition of the belt grain conveyor under the scenario of changing operating conditions, adaptive control of the driving force is extremely necessary, and for this reason, the research on the automation control technology of the double-drum variable-frequency driving force of the belt grain conveyor is proposed. The double-drum inverter drive model of belt grain conveyor containing motor, reducer and drum is constructed, and the unbalance problem of double-drum inverter drive of belt grain conveyor is analyzed in depth under the change of operating conditions in combination with the model; in the control stage, in combination with the principle of mechanics, according to the relationship between the traction force and the speed of the drum as well as the drive imbalance, the power supply frequency of the corresponding motor is controlled adaptively to achieve the stability of the belt grain conveyor. The power supply frequency of the corresponding motor is adapted according to the relationship between the traction force and the speed of the drum and the driving imbalance state, so as to achieve the purpose of stable operation of the belt conveyor. In the test results, the change of load intensity per unit area of the conveyor belt did not lead to abnormal fluctuation of the actual transmission speed, and it always operated stably according to the rated transmission speed.

In this study, the essential characteristics of momentum in tennis matches and its impact on the match results are discussed in depth by incorporating automatic control principles. By analyzing data from the 2023 Wimbledon Tennis Championships, we developed a statistical model of momentum which incorporates a feedback mechanism into the automatic control principle to improve the accuracy and responsiveness of predictions by automatically adjusting the momentum assessment parameters through continuous monitoring of data changes during the match. By performing two randomisation tests, the model demonstrates that momentum has a significant effect on the outcome of the race, where the p-value of the alignment test is close to zero, which indicates that momentum varies beyond the range of random fluctuations. In addition, the model also shows excellent prediction ability and is able to achieve 99% classification prediction accuracy on the test set. Meanwhile, in this paper, Kalman filtering is introduced into the momentum statistical model to achieve real-time monitoring and accurate assessment of momentum changes in tennis matches. Afterwards, using concepts from automatic control theory, such as feedback control and predictive control analysis, we achieve realtime monitoring of the match data, using Model Predictive Control (MPC) algorithms, combined with time series analysis and machine learning methods, to predict the trend of momentum during the match. Through this interdisciplinary approach, this research not only improves the understanding of the impact of momentum in races, but also provides a new scientific basis and tools for the development of sports training and race strategies.

To address the issue of slow location speed in magnetic location systems for capsule endoscopes that are based on permanent magnet and sensor arrays, an FPGA-based magnetic locating system for capsule endoscopes is proposed. This system reads magnetic flux density data from permanent magnet collected by three-axis digital sensors using FPGA. The linear algorithm is written into the FPGA to calculate the location, and the results are displayed on LCD to achieve the precise location of the capsule endoscope. The experimental results show that the sensor array can be read using FPGA in just 40ms. The linear algorithm produces an average location error of 12.56mm and an average angular error of 8.19°. These results demonstrate that the system can accurately and rapidly locate capsule endoscopes.

This system utilizes STM32F103C8T6 microcontroller as the control core, which adopts timer output comparison function to realize operation control, and combines with PID algorithm to design and realize motion target control and automatic tracking system. It integrates various components such as OpenMV4 H7PLUS, DS3120 single-axis digital servo and MG995 servo motor equipped with two-dimensional electric gimbal, indicator key selection module, OLED LCD display, voltage reduction module, air model 3s lithium batteries and other modules. OpenMV samples the laser color thresholds, and calculates the coordinates of the center of the output laser point after preprocessing; This information is then displayed on the OLED screen and received by the MCU which feeds the gimbal to get the relationship between the servo angle and the horizontal and vertical lines of the target surface, and controls the rotation of the gimbal through the adjustment of the duty cycle of the PWM signal, as well as realizes the function of selecting the mode according to the built mode. Experimental results show that in the movement of the target control laser (red) process, the PTZ response is stable, to achieve the reset, target paper rotation test; automatic tracking system achieves the laser tracking function and meets the acoustic and optical prompts and reset braking; measurement inaccuracies are slight in absolute value; all indicators are better than the design requirements.

The sudden change in tension of the cable that bears the weight of its own cable and the weight of the submersible causes it to deform and expand during the winch transmission process, resulting in problems such as breakage, wear, bulging, and relaxation of the cable in the traction winch groove. This seriously restricts the development of deep-sea cable exploration technology in China. Therefore, this paper focuses on the problem of fracture and wear of cables in traction winch transmission. By using ultrasonic assisted electroplating to modify the surface properties of the cable groove, the influence of winch variable friction coefficient on the release of cable tension is studied.

With the continuous development of meta-universe technology, people have started to explore the teaching mode of "meta-universe + education". The new teaching mode of "three-teacher classroom" constructed by meta-universe is mainly composed of virtual digital human teachers, intelligent robot teachers and human teachers, which can realize the collaborative teaching of teachers in the virtual world and the real world and carry out classroom teaching simultaneously. Users can learn and explore in the virtual world, while receiving personalized tutoring in the real world to enhance learning. The educational meta-universe can obtain a large number of high-quality educational resources across the time and space constraints, and provide immersive learning environment and personalized teaching support, so as to reduce the pressure of human teachers, give full play to the role of human teachers in the physical and mental development and ability training of students, so as to promote innovation in the field of education and promote the high-quality development of education.

In order to optimize the overall parameters of semi-active suspension, a random road surface model and a dynamic model of semi-active suspension are established based on spatial geometric relationships. The LQG control strategy based on optimal theory is adopted to suppress the pitch motion of the vehicle body and optimize the performance parameters of the suspension. The ratio of the root mean square performance parameters of semi-active suspension and passive suspension is used as the fitness function of the particle swarm optimization algorithm, and a single level of road input is discarded, Design a road excitation model with mixed road surface levels to adapt to the high-speed driving of vehicles on different road surfaces. Adjust the weight coefficients of the fitness function to a certain extent, and finally iteratively optimize the LQG controller weight coefficients corresponding to different road surfaces to further suppress the vehicle's pitch angle and improve the safety and stability of vehicle driving.

Aiming at the problems of poor real-time performance and low level of informatization in marine diesel engine processing workshops, a monitoring system architecture for marine diesel engine processing workshops based on digital twins is proposed by combining digital twins with marine diesel engine processing workshops. On this basis, a digital twin workshop was constructed and functional modules were divided according to requirements. Key technologies such as physical workshop digital twin model construction, machine tool processing simulation, and machine tool status management were studied, and three-dimensional visual monitoring of marine diesel engine processing workshops was achieved. Finally, the effectiveness of the system was verified through the application of a certain marine diesel engine processing workshop.

Due to the widespread application of coupled van der Pol chaotic systems in the field of science, So the robust control problem of a class of coupled van der Pol chaotic system is studied. Firstly, the complex dynamic behavior of the system under certain parameter conditions is verified by drawing the Lyapunov exponent diagram, chaotic attractor diagram, and bifurcation diagram of the system with parameter changes. Secondly, the sliding mode control, high gain robust control method and high-frequency robust control method are used to design the controllers to control the chaotic system. Finally, numerical simulations are conducted to validate the effectiveness of the designed controllers.

This paper used the text mining approaches to investigate the Chinese color terms in Song-dynasty lyrics and to explore the history, chromatics, linguistics and folklore reflected by colors in the Song Dynasty. The colorful Song Dynasty was observed through the color terms in the lyrics. Using the Python programming language to mine the text, this paper 1) extracting and analyzing the frequencies of all colors in the Complete Song-Dynasty Lyrics. Red, yellow, and green were found to be the 3 highest-frequency hues. 2) explored Chinese color terms and the collocations of red and white through the method of dual-direction extracting. The color term frequency ranking bar charts and collocation frequency ranking lists were obtained. The feminine characteristics and Chinese characters’ evolution were revealed in the collocations of red. The informal and romantic language characteristics and the existence of dialect and fairy tales were revealed in the collocations of white. 3) explored the top 20 poets’ preferences for using the collocations of red and white, as well as the popularity of these collocations in the Song Dynasty through the method of word co-occurrence. This paper could provide quantitative analyses for further exploration of the Song-dynasty lyrics and color aesthetics.

Offline reinforcement learning is proposed to learn policy from static data. However, when the model is deployed to real world, models without safety constraints can cause unsafe problems in reality. Using negative rewards or restricting the agent's action space for unsafe policies in related work may lead to models being overly conservative or aggressive, thus failing to balance task performance and safety effectively. In this paper, we propose a safety offline reinforcement learning method based on knowledge constraints by adding prior expert knowledge from the static data set. First, our algorithm utilizes an expert model to evaluate safety checks on state-action pairs, aiming to ensure that the model learns from safe data. Second, we use adaptive adjustment factors to impose safety constraints on the current policy. When the current policy is evaluated unsafe, constraints on the policy are heightened. Conversely, when the policy is evaluated safety, the model optimizes task performance accordingly. Experiments demonstrate that our algorithm outperforms baseline algorithms in safety, exhibiting reduced variance and enhanced training stability.

The Deepfake technique utilizes GAN to alter facial features, but can lead to accidental modifications and imperfect details. We propose an improved CycleGAN model that addresses these issues by improving the generator architecture, optimizing the loss function, and introducing an age-constraint strategy. Experiments on the CelebA and IMDB-WIKI datasets demonstrate that our model generates more realistic in-depth faked gender images compared to traditional methods, with average content accuracies of 84.30% and 63.7% for male and female faces, and FID scores of 32.89 and 3.58, respectively.

Predicting postpartum suicide tendencies is crucial for maternal health and public safety. This study compares four machine learning models for predicting these tendencies: Particle Swarm Optimization with Random Forest (PSO-RF), Random Forest, Naive Bayes, and Logistic Regression. PSO-RF outperformed the other models, achieving an accuracy of 0.991, perfect precision, high recall, and an ROC_AUC of 0.999. Random Forest also performed well, but with a slightly lower recall. Naive Bayes and Logistic Regression had lower accuracy and ROC_AUC scores, indicating less reliable predictions. Overall, the study suggests that PSO-RF is the most effective model for predicting postpartum suicide tendencies.

Automatic guided vehicle (AGV) has important application value in the field of logistics and manufacturing. The stability and accuracy of its navigation system are crucial to realize intelligent production process. In this paper, an integrated navigation strategy combining laser ranging sensor and gyroscope positioning is used to realize the adaptive navigation of AGV along the wall in different environments. In the case of a wall, the distance from the wall is measured by a laser ranging sensor, and the position information and attitude Angle information of the AGV are obtained through operational processing. Through the double PID closed-loop control, the AGV position correction and the automatic calibration of gyroscope heading Angle are realized, and the navigation accuracy and stability are effectively improved. When the AGV encounters a wall-free area, the system automatically switches to gyroscopic inertial navigation mode to ensure navigation continuity and accuracy.

Models built with deep learning has recently shown great success for addressing the challenging task of single image superresolution (SR). In-depth analyses of the representative models (e.g., EDSR, RCAN etc.) indicate the state-of-the-art performance is achieved by not only the claimed contributions (e.g., network structure, training strategies and loss functions) but also some underlying tricks, which is usually underemphasized. Our experimental analyses reveal that the tricks usually have a significant impact on the model’s performance, sometimes may play a more important role over the core designs. To gain more insight into this and provide a comprehensive analysis for providing SR related user guideline for those who are interested in SR, we’ve collected a bag of super-resolution tricks applicable to various stages (i.e., data pre-processing, data augmentation, loss function, training procedure, model structure, and model inference) of model implementation, and carried out extensive experiments to investigate what tricks are significant and how much can they improve the results over the base line. Through extensive evaluation, effect of different tricks is identified and thus encapsulated to various strong base lines for obtaining better quantitative and qualitative results. More importantly, the task of real-world SR benefit greatly from our formulated tricks, which is demonstrated to improve the PSNR by more than 0.4dB.

With the rapid development of cognitive neuroscience technology, researchers are focusing on the performance and regulatory mechanisms of working memory in EEG activity increasingly. Working memory is a crucial link in the information processing of the human brain, involving the temporary storage and manipulation of information, and directly affecting learning, understanding, and decision-making abilities. Firstly, this article starts with the definition and role of working memory, introduces the application of EEG technology in the study of EEG working memory, provides an in-depth review of EEG research related to working memory, and explores the application of EEG technology in foreign language learning. Finally, a summary and outlook tend to provide reference for researchers in the fields of cognitive neuroscience and neuroeducation in working memory for foreign language learning.

Currently, design units manually design the computer interlocking upper machine interface in railway station signal layout diagrams and use AutoCAD software for drawing. In order to reduce the demand for manual operation, improve the reliability and accuracy of data, a computer interlocking upper machine interface image recognition method based on OpenCV is proposed. By quantifying the geometric characteristics of various types of signal equipment on the upper machine, this method can quickly and accurately extract the screen coordinates of signal equipment, enabling users to easily obtain an overview of the layout of signal devices in the station.

Modulation and demodulation of optical signals are important links in laser communication links. Due to the influence of atmospheric turbulence, light intensity fluctuations, beam drift and other phenomena will occur during the optical transmission process, which increases the difficulty of demodulation and cannot meet the low bit error rate requirements in complex atmospheric environments. This paper studies the atmospheric attenuation effect and atmospheric turbulence effect, and analyzes the turbulence channel model. On the basis of studying the four machine learning algorithm models of Deep Belief Network (DBN), Support Vector Machine (SVM) and the basic principles of PPM modulation, the DBNSVM was designed. Demodulation algorithm, and improvements were made to DBN algorithm parameters based on PPM signal characteristics. Experimental results show that the DBN-SVM demodulation algorithm can significantly improve the demodulation performance of the communication system. For 4-PPM signals, when the signal-to-noise ratio is 14dB, the symbol error rate of DBN-SVM demodulation is lower than 10⁻⁶.

In the late stage of oil field development, due to casing loss, the injection-production relationship is very imperfect, and the geological reserves are difficult to be used effectively. The standard formation of Nen-2 index bed is always the high occurrence of casing loss. In order to effectively guide the prevention and control of casing losses, based on logging interpretation technology and information technology, combined with the lithology and electrical characteristics of oil shale in Nen-2 index bed, the quantitative and qualitative evaluation methods of SP outliers of oil formation logging curves are given, and a set of automatic identification methods and application programs are established in Nen-2 index bed flooded waters to achieve the batch visualization technology effect of "thousands of Wells with the same map". It realizes the identification of abnormal formation pressure and prediction of casing damage area by using logging data, and provides guarantee for the prevention and control of high quality casing damage.

The number of remote sensing satellites is increasing, and the types and number of payload are also increasing. User demand is also increasing, and satellite resources are still relatively tight. To study how to coordinate the joint observation of multiple satellites and multiple payloads, give full play to the overall observation benefits of remote sensing satellite resources, and meet the increasingly complex task requirements of various users, all available resources must be integrated. Realize the overall management of remote sensing satellite resources and task collaborative planning, and improve the ability of data acquisition and support through the effective integration of various resources. Research on intelligent task planning technology based on multi satellite and multi load. Through the research and construction of satellite and ground resource models, and standardized requirements and task models, the constraints of intelligent satellite task planning are realized, and the optimal scheduling under conventional planning is realized. Research and build the multi-objective optimization planning model of resource observation to improve the intelligence of task planning algorithm. Scientifically arrange satellite imaging and give full play to the efficiency of satellite use.

A set of knowledge graph for intelligent design of power transmission project in the early stage is developed in this paper. Using multi-source data fusion, all kinds of data with different formats are cleaned and integrated, the data of different layers are verified with the actual mapping, and the power knowledge is integrated into the map. Thus, a set of efficient, fast, and intelligent auxiliary decision-making system for transmission site selection and line selection design is finally realized, which aims to provide the basis and paradigm for the development of intelligent design in the early stage of a new power system.

With the continuous development of science and technology, the application range of intelligent robots is becoming more and more extensive, and indoor robots have become one of the hot spots in research. How to make indoor robots accurately and effectively detect and avoid irregular objects and complete path planning tasks has become the key to research. Initially, a variety of sensors including lidar, ultrasonic sensors, and cameras, are used to obtain environmental information. Lidar is used to provide high-precision distance measurement data, ultrasonic sensors are used to detect obstacles at close range, and cameras are used to obtain visual information and identify complex environmental features. Then, the sensor fusion algorithm based on Kalman filter is used to synthesize these multi-source data, eliminate the redundancy and inconsistency between the sensor data, and form a more accurate and comprehensive environment model. In the path planning stage, the improved A* algorithm is used to preliminarily plan the global path to ensure the optimality and feasibility of the path. Then, the Dijkstra algorithm was used to refine and optimize the local path to improve the smoothness and security of the path. In order to achieve real-time obstacle avoidance, dynamic adjustment is carried out by combining the dynamic window method (DWA) and the artificial potential field method (APF). The DWA method calculates the optimal control command in real time according to the current motion state to ensure the flexibility and response speed of the robot in complex environments. The APF method further improves the robustness of obstacle avoidance by introducing a virtual potential field to guide the robot away from the obstacle and move towards the target point. From our extensive experimental results, we can observe that the proposed method significantly improves the performance of the robot in obstacle avoidance and path optimization.

As the global economy and shipping industry continue to expand, transportation vessels are increasingly moving towards larger scales. This shift necessitates enhanced control performance and maneuverability for ships. This paper aims to bridge research gaps in the domain of collision risk management based on maritime rules, specifically focusing on intelligent collision avoidance strategies for underactuated merchant ships. The research integrates advancements by employing the MMG separation modeling and examining various algorithms to enhance autonomous navigation. By combining the improved artificial potential field method with the ship domain model, this study conducts ship simulation navigation experiments under different operational conditions. The primary objective is to develop a robust collision avoidance system that adheres to international maritime collision avoidance regulations, thereby increasing safety and efficiency in maritime operations. Future research will continue to refine these models and explore real-world applications to improve the reliability and practicality of collision management systems for commercial ships.

With the advent of the era of big data, the railway industry urgently needs to adopt efficient and safe distributed storage solutions. However, traditional distributed storage systems are insufficient in terms of data security and compliance. To this end, this article proposes a distributed storage plan based on the national commercial password algorithm (national secret algorithm), based on the Hadoop distributed storage system, and conducts national secret technology transformation. First of all, the SM4 countries secretly plus algorithm, and optimized the flow-type decryption operation during the read and write process, realized the parallel decryption of the data block, greatly improving the transmission efficiency of large files. Secondly, transform the KeyProvider API interface, integrate the national secret key management system, so that it supports the national secret algorithm and implement the key management of the key. In addition, the core functions such as KERBEROS security identity certification and role -based authorization are implemented. This scheme ensures that data can be stored efficiently and safely through the addition of unreasonable storage through the algorithm, which significantly improves the security of data storage, meets the requirements of the national confidentiality degradation, and provides a safe, controllable, high -efficiency and compliance solution for big data storage in the railway industry.

We present a novel approach for estimating confidence intervals (CIs) of bit error rates (BERs) in high-speed serial links, addressing the shortcomings of the conventional binomial method, which inaccurately presumes errors to be independent and identically distributed. Next, we introduce an alternate method which computes the CI from the Markov chain model of decision-feedback-equalization-induced error propagation in high-speed serial links, illustrating the difference of error distributions obtained by this method and the conventional approach. Finally, we propose a technique that curtails the computation time of our introduced Markov-chain-based method from quadratic growth to a more tractable level, thereby making the proposed method more practical when estimating CIs for low-BER systems. This method has the potential for enhanced accuracy in BER estimation for high-speed serial links.

This letter presents a multi-UAV task assignment method based on a Modified Sparrow Search Algorithm (MSSA). Initially, the research develops a UAV task allocation model for rescue contexts, incorporating constraints related to time windows. Moreover, the algorithm's flexibility is bolstered by incorporating dynamic adaptive weights and adaptive step size factors during the discovery and follower position update phases of the sparrow search. To circumvent the issue of local optima, a Gaussian-Cauchy mutation strategy are employed for optimization. Experimental results indicate that, compared with other improved sparrow search algorithms, the task assignment cost of UAV based on MSSA is significantly reduced.

This study utilizes the precipitable water vapor (PWV) data retrieved from four Global Navigation Satellite System (GNSS) stations in the Beijing area in 2023, along with related meteorological data. A rolling prediction model based on the Random Forest (RF) algorithm was proposed, effectively predicting rainfall and the rainstorm event on “July 31” in Beijing. The PWV values retrieved by GNSS were compared with those from the fifth-generation ECMWF reanalysis (ERA5) model, showcasing high consistency with an inversion accuracy of 2~3mm for GNSS. The rolling prediction method achieved an accuracy of 99% in predicting non-rainy events, a true positive rate of 92.25% in actual rainfall events, with a false alarm rate of only 1% and a miss rate of 7.75%, demonstrating the model’s efficiency in capturing rainfall events. Furthermore, during the rainstorm event on “July 31” in Beijing, the Root Mean Squared Error (RMSE) of the predicted precipitation stood at 0.81 mm, and the average coefficient of determination (R2) between true and predicted precipitation was 84.4%, with predictions highly consistent with actual rainfall trends and accurately predicting rainfall peaks. These results underscore the model’s potential as a valuable tool for rainfall forecasting, providing significant theoretical and practical support for predicting future rainfall events.

This study is based on Convolutional Neural Networks for intelligent classification of ship waste, efficiently categorizing ship waste into nine classes: A-plastic wastes, B-food wastes, C-domestic wastes, D-cooking oil, E-incinerator ashes, F-operating wastes, G-animal carcasses, H-fishing gear, I-electronic waste. Building AlexNet, VGG16, ResNet-50, and Inception-V3 Convolutional Neural Network models on the MATLAB platform and assess them on a custom dataset. The results indicate that Inception-V3 performed the best with a classification accuracy of 95.1%. Additionally, this study utilizes a Weighted Cross-Entropy Loss function to optimize the model, suppressing distortion caused by imbalanced training sets, resulting in an overall accuracy of 95.3%.

In order to solve the problem of how to effectively optimize the topological structure in complex systems, improve the accuracy and reliability of decision-making, and reduce the false alarm rate and false alarm rate, an intelligent topological structure decision-making optimization method based on knowledge map enhancement algorithm is proposed. Build an algorithm to enhance the knowledge map of intelligent topology decision-making, on this basis, build an intelligent topology and analyze its characteristics; According to the output results of intelligent topology, the decision-making optimization process of intelligent topology is designed. The experimental results show that after the optimization of this method, the storage space usage of each node has changed significantly, the accuracy of intelligent topology decision-making has been improved obviously, and the false alarm rate and false alarm rate have been greatly reduced, which has a good optimization effect.

In order to cope with the rapid transformation of energy demand and improve energy efficiency, as a new type of smart grid, Hebei South Power Grid urgently needs to conduct long-term adaptability analysis and regulation strategy research to ensure the efficient, reliable, and sustainable development of the power grid. Taking the power system as the research object, the genetic algorithm-based power system scheduling problem is studied in depth. Simulation experiments have been conducted to demonstrate the feasibility and superiority of this method. This article is based on an improved particle swarm optimization algorithm to train machine learning models and apply them to smart grids. In the exploration of energy loss costs, 150 million yuan was reduced to 120 million yuan, a decrease of 30 million yuan. This means that optimization strategies can effectively reduce energy loss in transportation and distribution, and improve energy utilization efficiency. This article provides a scientific regulatory strategy for the sustainable development of the southern power grid in Hebei, which has important theoretical and practical value.

Aiming at the problem that the traditional extraction method is not high in the accuracy of transmission line extraction under complex terrain, this paper proposes an improved classification extraction method to achieve effective transmission line extraction. In this paper, on the basis of KDTree algorithm clustering, the cloth simulation filter is added to make the classification accuracy higher, the operation speed is faster, and the interference of ground surface extraction of power lines is reduced. The extracted power line point cloud is again clustered, and each line is classified and given color. In order to verify the effectiveness of the proposed method, an experiment is carried out on a self-collected point cloud data set of transmission lines. The experimental results show that the improved algorithm can perform the task of transmission line classification well and achieve the expected effect.

Garbage classification involves the sorting and separation of waste to efficiently recover resources and minimize negative environmental impact. Currently, this process heavily relies on manual labor. In this study, we propose a lightweight residual network (LR-Net) to automate garbage classification. Initially, we introduce heterogeneous kernel-based convolutions (HetConv) and integrate them with residual connections, forming the lightweight residual block (LR-Block). Subsequently, we enhance ResNet18 by replacing its residual blocks with LR-Blocks, thus constructing LR-Net. To comprehensively validate LR-Net's effectiveness, we conduct comparative experiments on two open-source garbage datasets (TrashNet and Garbage dataset). The experimental results demonstrate that LR-Net achieves superior classification performance, with accuracies of 88.92% and 90.25% on TrashNet and Garbage dataset, respectively.

High-fidelity digital twin models of physical entities are the basis for the efficient assembly. Hull components have large scale, large quality, and high complexity characteristics. However, existing reconstruction methods of ship components model has low modeling accuracy, poor timeliness, and poor versatility, which has seriously affected the efficiency and quality of ship assembly. Therefore, combining feature point cloud recognition and feature stitching technology, a new fast reconstruction method for large-scale parts is proposed. Firstly, by analyzing the assembly process of parts, the constraint features of the assembly process are determined. Using the theoretical assembly model for training, an improved deep learning network for recognition and segmentation is obtained. Finally, based on the feature stitching technology, the large-scale model reconstruction is completed. The result shows the reconstruction speed of proposed method is significantly improved and the model has better accuracy, which provides a new way for the modeling of large-scale parts.

A system for elevator fault recognition based on open dataset construction is proposed in response to the frequent occurrence of elevator accidents and the inefficiency and timeliness of maintenance and inspection methods. The system involves elevator Internet of Things (IOT) and computer software technology, including the elevator fault recognition front end, server end, and client end. The data acquisition device of the elevator fault recognition front end collects video stream data and sends it to the edge computing device, and the fault algorithm model library of the edge computing device including fault algorithm models, and the fault recognition module will use the data to recognize faults in the video stream. When maintenance engineers detect the elevator operation fault from the video stream data, they input the fault characteristics into the model training server through the client’s model training app. The model training server, based on transfer learning, will apply the input fault characteristics to train new fault algorithm models and update them to the fault algorithm model library. This system provides maintenance engineers with open elevator fault model training services, continuously updating the model dataset to improve fault recognition accuracy.

This article designs a cognitive data collection paradigm that integrates PVT (Psychomotor Vigilance Task) with cluster missions to activate operator mental fatigue states and accurately collect sample data. Based on the sample data, a support vector machine (SVM) based classification model for the mental fatigue states of cluster drone operators is trained and constructed. Using this model, a cluster drone operation duration determination model is developed by setting criteria for extreme fatigue states and calibrating model thresholds. Through 100 rounds of 5-fold cross-validation on the dataset, the results indicate an accuracy rate of 95.5% for determining the mental fatigue states of cluster drones.

With the advancement of autonomous driving and robotic technologies, the demand for precise navigation and localization is increasing. To address the problem of error accumulation in inertial navigation and the challenges in mining data from inertial measurement units (IMUs), this paper proposes a velocity estimation network based on IMU arrays. This network integrates methods from self-supervised and supervised learning to enhance the accuracy and robustness of velocity estimation. We designed a network structure comprising an encoder and a velocity decoder. The encoder utilizes an improved BERT model for self-supervised pre-training, employing masking reconstruction techniques to effectively extract features from IMU data. The velocity decoder adopts a multilayer perceptron (MLP) architecture to further integrate features for accurate velocity estimation. Experimental results validate the effectiveness of the proposed method, demonstrating that it can effectively utilize data from IMU arrays to achieve accurate velocity estimates.

In the realm of computer vision, pedestrian detection is a pivotal and indispensable task with holding significant relevance for a multitude of applications. The progression in general object detection has propelled two-stage pedestrian detection algorithms to the forefront of performance. However, current works predominantly concentrate on overcoming occlusion challenges, while certain intrinsic constraints within the two-stage framework remain overlooked. This study identifies a feature mismatch issue within the pedestrian region proposal extraction module as a pivotal factor affecting the detection efficacy of two-stage pedestrian detectors. Therefore, we propose a receptive-field adaptive region proposal network and it harnesses the power of multi-receptive field features and deformable convolution, thereby mitigating the feature mismatch. Experimental results suggest that our proposed region proposal network markedly enhances the performance of two-stage pedestrian detection algorithms.

The task of role-oriented dialog summarization aims at generating appropriate summaries for the different roles in a dialogue. Although existing approaches are able to handle the dialogue content of roles, the extent of the contribution of different roles in a topic has not been analyzed in depth. In this paper, we propose a new model, RoleAdapt, which aims to balance the contributions of each role. RoleAdapt determines the type of summaries that need to be generated by recognizing the Prompt, efficiently identifies the content that contributes more to the topic, and dynamically adjusts the token-level vector representation, which is followed by a decoder that generates the summaries for each role. This approach not only evaluates the contribution of each actor, but also ensures the integrity of the contextual content. We validate RoleAdapt on two publicly available benchmark datasets, CSDS and MC. Experimental results show that the model achieves significant progress on both datasets. Further analysis shows that balancing role contributions is crucial for generating correctly structured summaries.

There are two agents A and B with each having their own jobs. One agent wants to minimize the total weighted completion time. Another agent wants to minimize the total completion time. Some structural properties of Pareto optimal solutions are given and a bigger WSPT scheduling set is found which contains all Pareto optimal solutions. In the WSPT scheduling set, we design an algorithm in polynomial time to get total KS-fair schedules while it proves that the upper bound of PoF value equals a half. Moreover, the above algorithm can be modified to get the KS-fair schedules in Pareto optimal schedule set, and prove that the bound of PoF is exactly a half.

Conventional phase-shift method can obtain the topography information of the object surface by projecting phase-shift fringe sequence and capturing deformed pictures. However, overexposure on shiny surfaces causes the loss of fringe information received by the camera, thus reducing the measurement accuracy and integrity. To solve this problem, an adaptive fringe projection method based on speckle-assisted fringes and mask fusion is proposed. Firstly, the optimal projection intensities are obtained by fusing multiple mask images. Then, the pixel-by-pixel mapping relationship between the camera and projector is established by using speckle-assisted phase shift method under unsaturated condition. Finally, the adaptive intensity fringe is generated and projected by the constraint of the mapping relationship. The 3D reconstruction is performed by the speckle-assisted phase shift method. Experimental results show that the RMS phase error of the proposed method is reduced by 94%, 33% and 12% compared with the contrast methods. The proposed method can accurately match the phase information of the overexposed surface and reconstruct objects with large surface fluctuations, thereby effectively solving the problem of 3D reconstruction of high reflective objects.

This paper explores vehicle routing optimization for energy efficiency in ITS, emphasizing platoon driving's potential to cut fuel use and emissions. It summarizes VRP's evolution, emphasizing fuel efficiency. A multi-objective model, considering energy, time windows, and platoon merging, is proposed to minimize fuel consumption. It factors in vehicle speed, loads, customer time windows, and platoon feasibility. Validation using genetic and particle swarm algorithms in MATLAB shows platoon routing can significantly cut fuel use and emissions, supporting ITS development.

Due to its capacities in improving spectrum efficiency, reducing hardware cost, and accelerating signal processing, the integrated sensing and communication (ISAC) has been widely discussed and investigated in recent years. In this paper the secure downlink transmission of a unmanned aerial vehicle (UAV) communication network is considered, in which the terrestrial base station performs ISAC to sensing multiple targets. Aiming at maximizing the average secrecy rate (ASR) while ensuring the sensing performance is higher that a specific threshold, the trajectory of the legitimate UAV and the transmit beamforming of the base station are jointly optimized. Since the considered ASR maximization problem is non-convex, we resort to the state-of-the-art deep reinforcement learning (DRL) based algorithms. Firstly the optimization problem is transformed into a Markov decision process (MDP), then the UAV trajectory as well as the base station beamforming are jointly optimized. Simulation results demonstrate the effectiveness of the proposed algorithm and validate the superiority compared with benchmark methods.

To address the challenge of robotic scrub nurse not achieving the standard "hand-to-hand" delivery of surgical instruments, this study introduces a robot-assisted instrument delivery method specifically designed for the field of oral medicine, employing composite vision guidance to enhance accuracy and efficiency. We construct three-dimensional delivery models for types Q, N, and Z of oral instruments and establish a composite vision system comprising RealSense D435i (eye to hand) and Leap Motion (eye in hand). The global space employs an improved YOLO v5 algorithm to identify and approximately locate the target instrument-receiving hand, while the local space leverages Leap Motion for more precise posture detection and tracking. Utilizing multi-dimensional posture data from the local vision and the type of instrument to be delivered, the system calculates the instrument delivery posture and plans the delivery path. This method has been experimentally validated on an oral instrument automatic delivery system set up in our laboratory. The results demonstrate that the average error of the delivery posture meets the actual requirements, effectively achieving standardized and precise "hand-to-hand" delivery of various types of oral instruments.

In response to the need for controlling unauthorized electrical appliances, we have developed a deep learning-based smart appliance recognition system. This paper proposes an innovative approach that effectively extracts features of appliance usage current, such as instantaneous fluctuations, peak currents, and transient changes caused by the starting or shutting down of appliances, using the Transformer's encoding layer. We refined the multi-head attention mechanism and removed the decoding module to simplify the model and focus on feature extraction. After the feature extraction phase, we employed ResNet as the feature mapping module, in conjunction with the enhanced Transformer. This model surpasses all previous models in identifying incremental current produced by appliances and is capable of adapting to the recognition of a large scale of appliances.

Partial Label Learning (PLL) represents a common multi-class weakly supervised learning challenge, where each instance is linked with a set of possible labels, only one of which is the ground-truth label, and it remains unknown during the training. A common strategy for solving this multi-class problems is to decompose it into a series of binary classification problems based on certain coding rules (e.g., OvO, ECOC), and then process each binary classification problem separately using the original features. To address the deficiency of discriminative information in the original features, a feature enhancement mechanism is proposed to introduce label-specific features to improve the discriminative performance of features. Firstly, label-specific features are generated for each decomposed binary classification problem, and then fused with the original features. The fused features are used to design corresponding binary classifiers. Additionally, due to the presence of a large number of false positive labels in the training examples, a semi-disambiguation strategy can be employed to remove some false positive labels, thereby "slimming down" the candidate label set as a preprocessing step to reduce its ambiguity. The proposed partial label learning algorithm based on features enhancement and semi-disambiguation proves its effectiveness on synthetic and real-world data sets through extensive experiments.

The imaging of root canal is one of the key technologies in the process of robot root canal preparation. Therefore, a method of root canal imaging based on active vision was designed. First, combine the detection accuracy index, a comprehensive evaluation value based on pixel proportion and resolution was proposed to judge the quality of root canal images.Then, the area size of the focusing window at different imaging positions is changed according to the distance of the lens feed, so as to reduce the visual interference caused by other backgrounds and achieve the effect of automatically capturing the position of the root canal. Finally, an autofocus strategy from coarse to fine was used to find the best imaging position of the root canal by single step search. An active vision system composed of a robotic arm, an operating table, a global depth camera and a terminal endoscope camera was built to carry out experimental research. The experiments showed that this method had good adaptability to light interference, and the obtained root canal images were very detailed, which could reflect rich information such as the shape, quantity and distribution of the root canal, and meet the accuracy requirements of the robot root canal preparation.

Wearable sensor-based activity recognition refers to the utilization of sensors worn on the body to collect human motion data. Through deep learning methods, different activities can be classified and recognized. This technology has wide applications in areas like health monitoring, smartwatches, and sports tracking. While deep learning-based Human Activity Recognition (HAR) has made great progress, challenges such as insufficient feature representation and underutilization of spatial and channel information still exist, leading to potential misclassifications. To address these issues, we propose an Attention-based Long Short-Term Memory Convolutional Neural Network framework (ACL-CNN). In this framework, the Convolutional Block Attention Module (CBAM) applies adaptive spatial and channel attention to the extracted features, highlighting important spatial positions and channel correlations to better capture behavioral characteristics. Additionally, the Long Short-Term Memory (LSTM) module processes the time-series data after feature extraction, effectively capturing long-term dependencies in the temporal sequence and enhancing the model's capability to model time-series data. The experiments are carried out on two publicly accessible datasets, namely WISDM and OPPORTUNITY. The results show that our method achieves the best performance on all four evaluation metrics. This validates the effectiveness and superiority of our method in wearable sensor-based activity recognition.

Maintaining the health of rolling bearing is crucial for ensuring the safe operation of rotating machinery system. In order to accurately and effectively monitor the condition of rolling bearing, a fault pattern recognition method based on the fusion of dominant features is proposed. Initially, the method involves performing wavelet packet decomposition on bearing vibration signals to extract multiple frequency domain features. Subsequently, the class separability of each frequency domain feature is assessed, and the dominant features are identified. These dominant features are then combined through dimension reduction to obtain comprehensive features. Finally, the comprehensive features are used to distinguish normal bearing signals from various faulty bearing signals through classification. To assess the efficacy of the proposed approach, experiments were conducted utilizing the MFS rolling bearing dataset. The findings from these experiments demonstrate that the method achieves high accuracy in classifying faults of rolling bearing.

In this paper, we propose an uncertainty-aware map-level point cloud semantic segmentation method. We divide the point cloud map into multiple small blocks by a specific strategy, which serve as the input of an uncertainty-aware backbone. The backbone can predicts both the semantic segmentation label and the uncertainty of the prediction. During training, the uncertainty-aware backbone first extracts the high-level features of point cloud through an encoder. Then the high-level features are fed into the learnable normalizing-flow module to approximate the dirichlet prior distribution. Both the parameters of the encoder and the normalizing-flow module can be updated during gradient backpropogation. During inference, the backbone predicts the dirichlet distribution parameter, which can be used for joint computation of semantic segmentation label and the uncertainty of the prediction. Experimental results show that our method has good performance in semantic segmentation and uncertainty prediction for unseen samples.

Entity-missing is one of challenges that the applications are facing, and fortunately, knowledge graph completion offers a solution to complete the missing one in a triplet when giving one entity and their relations. However, most of previous methods merely consider structural features or semantic information when predicting the latent links, which inevitably results in ignoring some critical information. Therefore, an approach combining the structure features and semantic information is proposed in this paper. Firstly, Tinybert is introduced as the encoder to extract semantic information from text description sequences, which can capture the hidden semantics from the text-description of entities or relations with fewer parameters compared with those large pre-trained language models. And then, a simple convolutional neural network is adopted to obtain structural features from triplets. Due to training the entity classifier with the two characteristics, our model outperforms those approaches that only considers semantic information while having a smaller size and faster speed than text-based methods with large pre-trained language models. Experiments on WN18RR dataset show that our approach achieves competitive performance compared to methods that only apply semantic information or structural features.

This study proposes a traffic flow prediction method based on graph attention network to address the needs of urban traffic congestion and planning. Our method first constructs a city traffic network graph, with road segments, intersections, and other traffic elements as nodes, considering traffic correlations and spatial structure to establish relationships between nodes. Then, utilizing the graph attention mechanism, it effectively captures the transmission of traffic information between nodes and the degree of correlation, thereby more accurately predicting future traffic flow. Experimental results demonstrate that our method achieved significant performance improvements in traffic prediction on real traffic datasets, proving its effectiveness and feasibility in traffic flow prediction tasks. This study provides new insights and methods for urban traffic management and planning.

In numerous tennis matches, people have discovered that there exists a kind of momentum in tennis competitions. When this kind of momentum forms, the player occupying the momentum exhibits an almost overwhelming consecutive win in this set. To study this kind of momentum, this paper uses the tennis match data of Jeff Sackmann on GitHub. Firstly, it conducts an exploratory analysis on the data of the men's singles tennis match at Wimbledon, and uses the TOPSIS comprehensive evaluation model with critical weighting to score the performance of the momentum of the players in a game, obtaining the broken-line graph of the momentum performance score for each point of both players. Afterwards, the random forest machine learning model with particle swarm hyperparameter optimization is used to fit and predict the change of momentum, and the confusion matrix indicates that the accuracy rate reaches 93.3%. The fitting results show that the number of times the player breaks serve and the cumulative number of game wins have outstanding contributions to the player's momentum.

This research proposes a fast Fourier transform-digital beamforming (FFT-DBF) based strategy for water level measurements on a bridge environment, utilizing linear frequency modulated continuous wave (LFMCW) radar. This paper presents two factors that affect the accuracy of LFMCW ranging on the bridge environment. By introducing a digital beamforming (DBF) algorithm, the waves from the undesired directions (UDs) are filtered. Moreover, the waves reflected by the strong stationary reflectors (SSRs) are filtered under the two-dimensional fast Fourier transform (2D-FFT) algorithm and the specialized cell averaging constant false alarm rate detector(CA-CFAR) algorithm. By using DBF, the main beam is able to point to the DD on the beam direction diagram. Simulation results show that the waves reflected from SSRs can be easily distinguished from the waves reflected by the water after processing the 2D-FFT and the CA-CFAR algorithm. More specifically, the results show that the estimation accuracy and stability can be further improved by using the FFT-DBF based strategy.

The study of composite scattering characteristics from target and environment is of great significance in the fields of target detection and identification in complex backgrounds. In this paper, typical environments such as canyons, earth-sea junctions, and islands are modelled by combining a composite environment modelling approach on top of the traditional stochastic rough surface generation. Subsequently, the multiple coupled scattering paths of the target and environment are analysed on the basis of the target and environment scattering calculation, with the objective of achieving the electromagnetic scattering calculation of the target and the composite environment. The numerical results demonstrate that the composite scattering model proposed in this paper meets the expected results in terms of solution accuracy and computational efficiency.

Hot waxy crude oil pipelines are prone to wax deposition formation during operation, which reduces the flow area of the pipeline and may lead to wax plugging and other accidents. The existing method of calculating wax deposition thickness based on pipeline operation data is difficult to obtain the distribution pattern of wax deposition along the pipeline mileage. In this article, in view of the large amount of information contained in the pipeline dynamic operation process, the idea of “time for space” is put forward. A method of calculating the distribution law of wax deposition along the pipeline is established by integrating the waxy mechanism model and data during the pipeline dynamic operation. The method converts the wax deposition calculation problem into an optimization problem with the dynamic operating hydraulic characteristics of the pipeline as the constraints, the internal diameter of the microelement segment along the pipeline as the decision variables, and the calculation results of the mechanism model and the information related to data during the pipeline dynamic operation as the objective function, and further rationally simplifies it into a linear programming problem. In this paper, through the analysis of arithmetic examples, it is proved that the method can realize the calculation of the wax deposition along the pipeline.

Automatic detection of code defects is crucial in modern software development. Traditional methods rely on manual feature extraction and rule matching, posing challenges in addressing intricate and evolving defect types. Deep learningdriven methods for software defect detection harness neural networks to extract defect characteristics from code, thereby augmenting both the efficiency and accuracy of detection. For deep learning tasks, both data preprocessing and network model design play equally crucial roles, directly impacting the model's final prediction outcomes. Nonetheless, prevailing research predominantly focuses on refining network architectures, often neglecting advancements in data preprocessing methodologies. To yield more precise data preprocessing outcomes, we propose a multi-stage source code structuring methodology. Initially, distractions within the code text are eliminated through deduplication and normalization procedures, followed by converting the text into a standardized fixed-length vector representation using disambiguation and vectorization techniques. Drawing upon the structured representation of the source code, we employ deep neural networks to discern defects within the code. Experimental findings demonstrate the pronounced superiority of the proposed approach over traditional methods on public datasets, with notable advancements observed in precision, recall, and F1 scores. Our study highlights the vast potential of adopting structured source code representations for enhancing the efficacy of software defect detection systems.

In order to address the challenge of inland water surface object detection, where small objects account for a large proportion and are prone to detection failures, limited GPU resources on embedded platforms, and deployment difficulties in unmanned surface vehicle (USV) object detection tasks, an improved lightweight YOLOv8 algorithm for surface object detection is proposed. Firstly, to mitigate noise interference and omission of small surface objects during downsampling in the backbone network, a Dilation-wise Residual (DWR) module is introduced to enhance the C2f module to efficiently acquire more contextual information and increase the model’s attention to small objects. In addition, given the complexity of the neck of the original network, a new feature fusion network, Multi-scale Screening-feature Pyramid Networks (MSFPN), is designed in this paper to reduce model complexity while maintaining accuracy. The experimental results demonstrate that the proposed algorithm achieves a 1.1% improvement in average accuracy (mAP0.5) for surface objects, with a 36.54% reduction in model parameters and a 14.81% reduction in computational load. It is proved that the proposed DM-YOLOv8 network model has better and more reliable performance in detecting the water surface of inland rivers and can meet the needs of safe autonomous navigation of USV with high real-time and flexibility requirements.

The multi-objective optimization scheduling of microgrids is crucial for reducing electricity usage, energy consumption, and environmental pollution. This paper proposes a microgrid optimization management method based on the improved Sand Cat Swarm Optimization (ISCSO) algorithm to enhance the economy, stability, and reliability of microgrid operations. By predicting distributed energy output and power load, an optimization objective is set focusing on operating costs and environmental management. The ISCSO algorithm optimizes energy output, outperforming the Grey Wolf Optimization (GWO) and Whale Optimization Algorithm (WOA) in scheduling. It effectively optimizes distributed power sources and power exchange with the distribution network, minimizing the microgrid's overall operating cost. The proposed method, model, and ISCSO algorithm are validated as rational and effective.

Modern blast furnace ironmaking technology primarily utilizes the thermal condition of the furnace belly to reflect the furnace temperature status. However, the complexity of the smelting process makes effective modeling and control extremely challenging. During the ironmaking process, the control of furnace temperature directly affects production efficiency and product quality. Since the furnace temperature is difficult to measure directly, the silicon content in molten iron is commonly used to reflect the thermal state of the blast furnace. Traditional methods for predicting the silicon content in molten iron have limitations and struggle to adapt to complex and variable production conditions. With the advancement of neural network technology, this paper constructs a prediction model for the silicon content in blast furnace molten iron by creating a hybrid of Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), and a Multi-Head Attention Mechanism (MA). Through deploying and analyzing the CNN-LSTM-MA model in a real production environment, the superiority of the CNN-LSTM-MA model in silicon content prediction has been verified.

Aiming at the path planning problem of unmanned aerial vehicles in low altitude flight in mountainous terrain, an improved genetic algorithm based on the idea of group parallel exchange is proposed. Firstly, by adopting a point wise planning approach, followed by a scatter fitting strategy, a smooth planning path can be obtained. In the experiment, the advantages of the proposed algorithm in terms of path length were demonstrated by comparing it with other algorithms. The results indicate that the proposed algorithm is not only theoretically feasible, but also achieves the expected results in simulation experiments, effectively solving the problem of the original genetic algorithm easily falling into local optima.

Timely detection and effective control of potential public health hazards are crucial for public health monitoring. This paper proposes a machine learning-based monitoring data anomaly detection system, which can efficiently handle largescale monitoring data from multiple heterogeneous sources. The system architecture adopts a distributed cloud computing design, integrating advanced algorithms such as Isolation Forests and Long Short-Term Memory Neural Networks, to fully explore the spatiotemporal features of the data, achieving precise detection and prediction of abnormal events and development trends. Large-scale experiments have verified the outstanding performance of this system on real influenza, pneumonia, and other monitoring data, significantly outperforming traditional statistical methods, and providing strong support for public health decision-making.

The Chinese Library Classification (CLC) number is a crucial identifier for indexing books in libraries, and accurate, efficient, and convenient book classification can greatly enhance the management efficiency of libraries. Addressing the issues of low efficiency and insufficient accuracy in current manual book classification and traditional machine learning methods, this paper proposes a book classification algorithm based on the BERT-TextCNN model to achieve automatic book classification. By integrating TextCNN into the BERT model, it can better capture local features and textual structure information. Experimental results show that the model achieves an average precision of 88%. Compared to benchmark models and traditional manual classification methods, the classification efficiency and accuracy have been significantly improved.

This article proposes an improved YOLOX model algorithm to improve the efficiency and accuracy of safety detection in power work sites, addressing safety risks such as improper wearing of safety equipment and environmental factors. This is crucial for preventing production accidents. This algorithm is optimized for three major challenges: small object detection, multi-target diversification, and complex backgrounds. By introducing shallow feature layer fusion, receptive field module RFB to enhance the ability of diverse object detection, and ECA Net attention mechanism to improve target differentiation in complex backgrounds, high-precision and real-time security detection has been achieved. The experimental results show that the new algorithm achieves mAP of 92.29% and detection speed of 36.06 FPS on a specially designed dataset, which is significantly better than the basic model, verifying the effectiveness and practicality of the algorithm.

A multiple UAVs gas pollution source localization method based on improved RRT algorithm is proposed to address the problem of gas pollution source localization in flood control protection areas, which is randomly distributed and difficult to monitor. Using FLUENT software to construct a CFD simulation concentration field, the improved RRT algorithm was validated, and the gas pollution source was successfully located, demonstrating its feasibility for implementation. Three algorithms were compared and analyzed in simulation environments with different wind speeds. The results showed that the improved RRT algorithm has good stability and robustness, effectively improving the success rate of UAV gas pollution source localization, shortening search time, and optimizing flight path. It provides an effective method for environmental monitoring in flood control protection areas.

To effectively address the Flexible Job Shop Scheduling Problem, an improved FOA is proposed, which incorporates dynamic parameters on the basis of the original algorithm to prevent it from getting trapped in local optima during the optimization process. The algorithm's global optimization capability and local optimization ability were validated through testing. Finally, relevant experiments were designed for the Flexible Job Shop Scheduling Problem, and the feasibility of the algorithm in solving the Flexible Job Shop Scheduling Problem was verified through experiments.

The issue of student poverty is not only a core concern in the field of education but also an important topic in socioeconomic development. Identifying and assisting impoverished students is crucial for achieving educational equity and social justice. However, the determination and classification of the poverty levels among students face numerous challenges due to various influencing factors. Therefore, this paper employs a k-means improved linearly separable support vector machine model to classify the levels of poverty. Experimental results show that the classification model performs well, with an accuracy of 97.419%.

Our study is based on the Long Short-Term Memory recurrent neural network (LSTM) model, aiming at mining the changing rules of consumption behavior features. Through feature extraction and descriptive statistics on the consumption data of 5415 students over a three-year period, we reveal the degree of discrete and changing patterns of consumption behavior. The model construction part introduces the principle and framework of LSTM and its advantages in capturing long sequence dependencies. The algorithm implementation part demonstrates the effect of predicting the consumption behavior in the fourth year, and the accuracy and robustness of the model are verified by the RMSE values of the training set and test set as well as the fitting situation. The results of predicting fluctuations in consumption behavior characteristics, including total daily consumption, frequency of consumption, and average consumption amount, are further analyzed. The results of the study help to understand the dynamics of consumption behavior and provide reliable data support for the determination of poor students, which can help to help poor students get out of difficulties more accurately.

This paper discusses the theory of embedded database SQLite and its application development on ARM platform, analyzes the characteristics of embedded database in detail, and expounds the architecture of embedded database on this basis. Based on the analysis of SQLite's architecture, powerful function, simple interface, fast speed, small space and other features, this paper introduces its development technology and the method of SQLite configuration and transplantation on ARM platform, and explains its application process with examples.

In the wind power industry, the algorithm library plays an important role in the development of control systems as a carrier for the application and accumulation of industry knowledge. In order to realize the development of wind power algorithm library, this paper analyzes the specific needs of the wind power industry for the algorithm library, combines the commonly used control system program design standards in the industrial control industry, proposes a design method of wind power algorithm library from the perspective of functional design and interface design, and describes the process of realizing it through IEC61131-3 language and C language. The algorithm library developed by this method has been applied to large-scale wind turbines.

The terrain awareness and warning system (TAWS) is an important part of the avionics integrated environmental monitoring system, it can issue warning to avoid dangerous terrain, avoiding controlled flight into terrain (CFIT) occurred in order to ensure flight safety. Study the important component of TAWS - the forward-looking alarm algorithm and composite view. The mathematical model of forewarning is established, and the composite visual display screen based on TAWS system is designed and realized. The forward looking alarm and synthetic visual display are verified by simulation. The results show that the forward-looking alarm algorithm is accurate, and integrated vision screen is clear. They can play an important role in improving flight safety in the congested airspace of the future.

Accurate forecasting of daily gas consumption of each station in a natural gas pipeline network is of great importance for natural gas companies to plan pipeline transportation capacity, formulate business plans, and carry out production scheduling and operation management. Daily natural gas consumption of gate stations is a complex time series data affected by factors such as season, cycle, and production plan. Challenges in forecasting daily gas consumption include the complexity of time-series data, the impact of external factors, and the need for effective integration of multiple features. Traditional single neural network models, such as Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN), struggle to adequately address these challenges. In this paper, a prediction model based on deep learning theory is proposed. The model is a combination of Long Short-Term Memory Network (LSTM) and Parallel Convolutional Neural Network (PCNN). The model uses PCNN to automatically extract the time series data features of natural gas gate station gas consumption and capture its changing trend; LSTM is used to capture the time series dependence of gas consumption. The proposed model aims to overcome the limitations of traditional models by integrating LSTM and PCNN to enhance feature extraction and capture complex dependencies in the data. The experimental results show that the proposed method has high prediction accuracy and stability, and can adapt to the complex and changeable pipeline network transportation environment, which provides strong support for natural gas pipeline network transportation management and optimization.

Currently, most of the encryption of offshore wind power information system uses the traditional Advanced Encryption Standard algorithm (AES algorithm). However, there are some problems with this method, such as poor security and stability, large computational workload, and the inability to effectively manage sensitive data. So we urgently need to improve this situation. In this paper, we propose a method to improve the traditional AES algorithm based on lightweight AES encryption algorithm. The method improves the implementation efficiency of the algorithm by reducing the number of cycles of the algorithm wheel function and using function encapsulation to define the operation form of the key expansion module. This scheme simplifies and optimizes the implementation process of the encryption algorithm while maintaining the security of the AES algorithm. The experimental analysis results show that the offshore wind power information encryption system based on lightweight AES encryption algorithm can improve the operational efficiency.

The rapid development of the global air transport industry makes the number of aircraft and flight traffic were on the rise, which led to tight airspace, and the air traffic control task is also increasingly heavy. Four-dimensional trajectory prediction algorithm is an important means to reduce air traffic congestion, improve airspace resource utilization and enhance airport operation efficiency. Based on the aircraft performance database, the parameters prediction and vertical track generation algorithm of 4D trajectory are given in this paper. The forecast information includes the altitude, speed and time at a certain flight way point, and the vertical profile is generated based on the forecast information combined with the selected optimization index. Finally, analysis and simulation are carried out to verify the effectiveness of the algorithm. It lays a foundation for solving flight track conflicts and optimizing flight scheduling.

Since China's low-orbit Internet satellites are still in the early stage of deployment, there is still a big gap in achieving global coverage. In view of the above problems, an intelligent decision-making planning algorithm for low-orbit satellites is proposed to achieve continuous coverage of low-orbit satellites in a certain area for a period, and to decide the globally lowest-cost task allocation scheme to guide the planning of low-orbit satellites and the decision-making of low-orbit satellites. The algorithm defines the cost calculation module, which calculates the cost of each satellite reaching each working position, generates a cost matrix and uses the classic Hungarian algorithm to calculate the globally lowest cost distribution scheme. To provide a continuous communication task as an example, an example of the application of the algorithm is given, which proves the correctness of the algorithm.

In response to the issues of basic Particle Swarm Optimization (PSO) being inclined to getting stuck in local bests, slow convergence precision, and low convergence rate, a Improved Chaotic Particle Swarm Optimization (ICPSO) based on hybrid strategies is proposed. This algorithm combines the adaptive inertia weight strategy, boundary symmetric mapping method, adaptive mutation idea, and chaotic optimization idea, enhancing the global search capability of the PSO, accelerating convergence rate, and increasing the likelihood of jump out of local bests. Finally, multiple classic test functions are picked to contrast the ICPSO with the basic PSO and other enhancement methods. The results indicate that the ICPSO converges faster, achieves higher convergence accuracy, and it is easier to converge to the global optimum.

With the frequent occurrence of natural disasters and emergencies, how to deliver emergency materials efficiently and accurately, is a key problem in emergency management. Traditional distribution path optimization method is often difficult to deal with complex and changeable emergency environment, therefore, this paper builds a emergency supplies distribution model based on the combination of trucks and drones, and puts forward a hybrid genetic algorithm to solve the model, realize the efficiency of emergency logistics distribution, and save the cost of emergency logistics distribution, provides strong support for emergency management.

An improved ant colony algorithm for path planning addresses issues like local optima traps, numerous inflection points, blind searches, and slow convergence. This method incorporates guidance from a potential field force to enhance effectiveness. Firstly, the blindness is reduced by canceling the obstacle repulsion and retaining only the target point gravity, and an improved potential field heuristic function is introduced to accelerate the convergence speed by adjusting the potential field constants and realizing the path planning under the target gravity. Secondly, dynamic pheromone volatility coefficients and state transition probabilities are introduced, and the pheromone of different quality paths is updated through the pheromone difference mechanism to realize the convergence of high-quality nodes, which solves the problem of local optimality. Finally, to reduce path turning points and enhance step length diversity, the pruning strategy is introduced to optimize the paths of the improved algorithm, and the paths are smoothed by using cubic B-spline curves. Simulation experimental results show that, Compared to several other algorithms, the improved algorithm demonstrates notable advancements in convergence speed, search efficiency, and final path length. These enhancements are highly valuable for practical applications in mobile robot path planning.

An estimation method for the direction of arrival (DOA) is proposed on the basis of a sparse representation method which tends to minimize the weighted variance. We relaxes the rigid constraints that the desirable sparse solution should have the minimum number of nonzero elements and provide a new diversity measure of sparsity which termed as soft sparsity. The often-used re-weighted iteration technique is exploited to find the new-defined sparse solution. More importantly, this method can find and estimate the DOA of the relatively weak sources. Experimental results validate the good performance of the proposed method.

As economic development and social progress bring convenience, they also bring a large number of problems. In the field of transport, the increase in the number of vehicles travelling on the road has led to an increase in traffic problems. With the introduction of intelligent transport, the problem of road traffic has been alleviated. in the field of traffic how to quickly and accurately identify the surrounding information to assist the driver in make decisions is crucial. In this paper, an improved YOLO algorithm model is proposed for solving related problems. In order to reduce the number of parameters in the model and accelerate the inference speed of the model, it is proposed to use Faster-Block as the backbone network, and further reduce the number of parameters by using GSConv convolution in the Neck part, and finally, the model is further compressed by using a pruning algorithm that learns convolution efficiently in a network thinning. And the improved algorithm was validated in the VOC dataset as well as the CCTSDB dataset for experiments, pruning the size of the model as well as the number of parameters while keeping the model mAP/0.5 without too much change, which reduces the inference time and achieves good results, and is more convenient to be used for deployment.

At present, the traditional recognition methods such as naked eye recognition and machine vision are mainly used for the surface defect image recognition of fan blades, and their robustness is poor. In order to improve the recognition accuracy of fan blade surface defects, an improved ResNet50 algorithm is proposed in this paper. 1416 fan blade images were captured by UAV, and the data set of fan blade surface defects was established through screening, classification and data expansion. In order to verify the validity of the experiment, the data sets of sand holes and oil pollution defects of fan blades were subsequently supplemented in this experiment, and the same data set processing method was used. With ResNet50 as the backbone feature extraction network, the channel attention mechanism SENet is introduced to enhance the reuse rate of useful features and suppress useless features. In order to reduce the risk of inactivation of some neurons using ReLu activation function, this paper uses Leaky-ReLU activation function to redesign the residual block. The experimental results show that the classification accuracy of the algorithm on the self-made data is up to the baseline model, which can meet the actual demand.

Intelligent logistics path information system based on improved genetic algorithm aims to improve the efficiencyandaccuracy of the algorithm by considering various constraints such as distance, and delivery time, etc. The systemalsouses Python programming language for data processing, operation and visualisation of the improved genetic algorithm, using logistics path data information to optimise the path planning of logistics vehicles to significantly improve transport efficiency and reduce logistics costs for the purpose of logistics enterprises. The system also uses Python programming language for data processing, operation and visualisation of the improved genetic algorithm to optimise the path planning of logistics vehicles by using logistics path data information, which substantially improves the transport efficiency and reduces the logistics cost purpose for logistics enterprises.

In this study, we propose a self-supervised method for learning universal neurocognitive representations from motor imagery electroencephalography (EEG) segments. Our model, named MAE-MI, is a masked autoencoder that learns robust and generic embeddings. The aim is to capture the essential features and recover the masked EEG signals during pretraining, rather than simply interpolating. Therefore, we design a connectivity guidance masking strategy. Furthermore, we optimize the encoder-decoder structure to suit the information density of EEG signals. We also introduce two finetuning modes for downstream tasks that are task-specific and subject-specific; they correspond to cross-subject and singlesubject evaluations, respectively. We assess the generalization performance of MAE-MI on a public motor imagery EEG dataset. The experimental results indicate that MAE-MI consistently outperforms the state-of-the-art methods with an average accuracy increase of 9.4% in single-subject prediction experiments; it also obtains competitive results in crosssubject experiments. We illustrate the feature extraction capability of MAE-MI by visualizing its reconstruction effects on corrupted EEG signals.

In this paper, the structure of the cyclone separator with an air volume of 15000 m³ /h is optimized, the structural data of the two-stage cyclone separator is determined by software, and the cutting particle size of the two-stage separator is determined by theoretical calculation, and finally the optimal structure is determined by the simulation of the flow field under different structures by using Fluent software.

Aiming at the problems of long planned path length, high repetition rate and long escape time, the bioincentive neural network algorithm is improved when it is applied to the full coverage path planning of the sweeping robot. Firstly, by modifying the activity value of the initial state of local grid neurons, and introducing the moving distance cost function to optimize the original direction function, the unnecessary turns of the sweeping robot will be reduced, and the planned path length will be shortened. Then when the sweeping robot enters the dead zone, the escape direction Angle is used to guide the sweeping robot to quickly escape the dead zone. Through the comparison and verification of simulation experiments, the improved algorithm can effectively reduce the path length, reduce the repetition rate, and enhance the escape ability while achieving full coverage, and the cleaning efficiency and quality of the sweeping robot are improved.

In order to optimize the effectiveness of power material scheduling, a periodic priority sorting recognition algorithm for power material scheduling is proposed. Normalized time series of power material dispatch instructions, clustering power material dispatch instructions based on periodicity; Consider the scheduling and sorting problem of electric power materials as the shortest time calculation problem, establish an objective function for scheduling and sorting electric power materials, and assign priority based on the importance of scheduling tasks; Based on the particle swarm optimization algorithm to identify the optimal periodic priority ranking results, a greedy algorithm feedback mechanism is introduced for priority feedback iteration to obtain the identification results of the priority ranking scheme for power material scheduling. The test results show that the application of this method has improved the utilization rate and task throughput of power material scheduling nodes, improved the fairness index of scheduling, and optimized the effectiveness of power material scheduling.

In order to solve the problem of power equipment fault, we can quickly find a set of alternative operation schemes with minimal impact on the stable operation of the power grid through graph computation, and propose the design of N-1 forethought fault analysis algorithm for power equipment oriented to graph computation. After the equipment fault, it is necessary to quickly assess its impact on the entire power grid; establish a fault model for the feature quantity and fault class to achieve the discretization of the sample, and N-1 predictive fault analysis of power equipment; use the algorithm design of N-1 predictive fault analysis to optimize the algorithm to find the optimal alternative operation scheme, which can be used to restore the normal operation of the power system. After testing, it is found that the method has a significant improvement in the speed and accuracy of power equipment fault analysis.

Currently, the main network operation equipment state estimation algorithms based on graph databases usually use machine learning or deep learning models for training and prediction. However, due to the diversity and complexity of main network operating devices, the generalization ability of the model is insufficient and it is difficult to adapt to the needs of different devices and scenarios. Therefore, this paper proposes a graph database-based state estimation algorithm for main network operating devices. Firstly, the spectrum-based graph convolutional network and graph database are constructed; secondly, the real-time monitoring and evaluation of the operating state of the equipment is carried out; then, by constructing the knowledge graph of the main network operating equipment; finally, by analyzing the historical data of the equipment, the failure mode and trend of the equipment can be found, so as to predict the future state of the equipment. The test results show that the graph database is highly scalable and can handle large-scale data sets. And the estimation algorithm of main network operation equipment state based on graph database has obvious advantages in fault feature extraction, computation time and modeling accuracy, and is able to detect the early faults of the equipment with an accuracy rate of 98.6%.

Aiming at the status quo that the defect detection method on the surface of the rollers is less efficient and unable to detect abnormal faults quickly in real time, a defect detection algorithm on the surface of the rollers based on improved DeepLabv3+ and migration learning is proposed. First, a lightweight MobileNetv2 network is used to replace the original backbone network Xception of DeepLabv3+ algorithm in order to reduce the number of model parameters and improve the inference speed; then the strategy of combining heterogeneous sensory field fusion and void depth separable convolution is used to improve the void space pyramid pooling structure to improve the information utilization and training efficiency of the model; finally, the attention mechanism is introduced to improve the performance of the model by using data augmentation and migration learning to improve the network training effect and enhance the accuracy of defect recognition on the surface of the rollers. The experimental results show that the improved DeepLabv3+ model has fewer iteration steps and faster training speed than the original model, the MIoU value has been improved by 3.46%, the number of model parameters has been compressed by 69.8% of the original model, and the segmentation effect on the defect edge has been improved significantly, which can complete the defect detection on the surface of the rollers in a fast and realtime manner, and has certain reference value and application significance.

With the continuous expansion of circuit scale, it is required that SRAM has good resistance to single event upset under different power voltages. Therefore, To investigate the effect of power voltage on multi cell upset of SRAM in FinFET process, simulation was conducted to investigate the effect of ions incidence at different positions on four cell FinFET SRAM under different power voltage conditions. Research has found that the combined effects of charge sharing and parasitic bipolar effects determine whether the SRAM cell upset. When ions are incident on the sensitive drain region, the multi cell upset is mainly determined by the charge sharing effect. As the power voltage increases, the ability of SRAM to resist single event upset is enhanced. When ions are incident into the well region, the drift charge collection mechanism of the sensitive drain region and the parasitic bipolar effect compete with each other to determine whether the cell upset. As the power voltage decreases, drift collection weakens, and the parasitic bipolar mechanism makes it easier for the cell to upset and recover. The results of this study have guiding significance for carrying out anti multi cell upset radiation reinforcement for FinFET SRAM.

In this paper, a monitoring biomimetic fish applied to aquaculture ponds is designed, which has the function of fixed-point cruise swimming and autonomous obstacle avoidance under water. It solves the problem that the aquaculture tank detector on the market is fixed, and the data of the aquaculture tank cannot be comprehensively and systematically monitored, achieves the effect of scientific monitoring, and fills the gap of aquaculture intelligence in our country. Based on Lagrange equation and D-H(Denauit-Hertenbery) coordinate transformation, the swimming dynamics model of bionic fish is established. BCF wave propulsion mode is adopted. The front part of the bionic fish body has a small amplitude, and the rear part of the tail fin generates forward thrust through swinging motion, which has high propulsion efficiency and propulsion speed. It can effectively reduce the stress response of fish. The influence of caudal fin aspect ratio on the hydrodynamic performance of the fish body was deeply studied, and the aspect ratio was calculated, and three sets of data were designed to select the more suitable caudal fin aspect ratio data. The motion mode of the bionic fish is calculated by the coupling of the envelope of the fish's amplitude and the sine curve, and the curve of the fish's body wave is drawn by MATLAB. CFD software FLUENT was used to conduct fluid simulation based on RANS equation, and UDF macro documents were used to define the tail wagging motion of bionic fish in the aquaculture tank. Based on the size of the bionic fish, the size of the fluid domain is extended outward by 50cm, and the mesh is divided. The velocity vector diagram, pressure distribution cloud diagram and velocity cloud diagram of the fish body under water flow movement are obtained. It is observed that the fish body is in a better state when the tail fin is selected.

Reinforcement binding is an important construction technology in construction and civil engineering. Currently, many construction processes in China still heavily rely on manual operation, with high work intensity, low efficiency, and high risk, making it difficult to achieve large-scale and industrial production. Reinforcement binding is a key component of reinforced concrete structures, which directly affects the continuity, stability, and safety of the structure. Therefore, this article proposes the CSW-YOLOv8s algorithm. Firstly, a Channel Feature Extractor (CFE) is designed, and a C2f-CFE module is constructed to enable the network to focus more on dense and homogeneous small-sized objects, solving the problem of difficult detection of small targets in the quality detection task of steel bar tying. At the same time, a Space Channel Feature Extraction Module (SCFEM) was designed to filter out interference information, enhance small target feature information, improve the problem of small target feature information being submerged during the convolution process, and solve the problem of target occlusion in the quality detection task of steel reinforcement binding. To address the issue of missed detection of individual steel bar binding points, Wise IOU is introduced as the loss function of this algorithm. The above improvements effectively solved the problem of insufficient accuracy in detecting small targets in YOLOv8s, and improved the detection accuracy of the model for steel bar binding points.

In recent years, deep learning has demonstrated powerful potential in air pollution prediction due to its outstanding learning ability, generalization capability, and handling of high-dimensional data. This article systematically reviews deep learning models for predicting air pollutant concentrations. Firstly, it examines traditional prediction methods and analyzes their advantages and limitations. Subsequently, it focuses on the development trajectory of deep learning in this field. Furthermore, it delves into models based on attention mechanisms, such as Transformer and Informer, which enhance prediction performance by adaptively focusing on key features. Additionally, the importance of data decomposition in prediction is emphasized. By decomposing complex data, deep learning models can be more effectively utilized to improve prediction accuracy. The article concludes by summarizing the current research achievements and shortcomings and providing outlooks on future research directions. This study aims to inspire relevant researchers and promote further development of deep learning in this field.

A set of azimuth swing mechanism was designed based on the slider-crank mechanism. Then, its structural combination has been described, and the working principle of the mechanism was analyzed. The inverse kinematics equation is obtained by the analytical method. The parametric virtual prototype model is established by Adams software, and the correctness of the model is verified. With the minimum pressure angle as the optimization objective, the optimized size and pressure angle of the mechanism are optimized, which improved the transmission performance of the mechanism. So, this paper provides a theoretical basis for the design, analysis and optimization of the same type of mechanism.

With the improvement of our country's requirements for the development of ecological civilization, as well as the further enhancement of our country's urban motorization level, highway sanitation is naturally in a critical positionintheconstruction of ecological civilization. Therefore, this study applies robotics-related technology to develop an intelligent garbage inspection robot that can intelligently avoid vehicles on highways and automatically identify and pickupgarbage. The robot uses STM32 microcontroller as the main controller, FPGA chip to identify and locate the garbage, and Raspberry Pi to realize the automatic navigation system, which applies ROS and deep learning technology tomakethe robot efficiently clean up and recycle the garbage, and at the same time, ensure the safety of the operationonthehighway. This study will provide certain value for the development of highway garbage cleaning robots.

Flexible sensing technology plays a key role in robotics intelligence, wearable devices, and biomedicine. Compared to rigid sensors, flexible sensors have advantages such as compliance, multimodal deformation sensing, stability, repeatability, hysteresis, and low cost. With technological advancements, flexible sensors will better adapt to large deformations in devices and provide reliable sensing. This article introduces a flexible monolithic multimodal sensor using liquid metal technology to detect stretching, twisting, bending, and pressing forces and spatial deformations through a single sensor. Unlike traditional resistive sensors, this sensor analyzes the rate of resistance changes in different parts to output signals. It features specialized liquid metal microchannels for various deformation modes, ensuring significant volume changes in corresponding microchannels while minimizing changes in other channels to prevent signal interference. By comparing the resistance changes in liquid metal-filled microchannels, the sensor can assess the extent of deformation in each channel accurately.

Photovoltaic cells play a crucial role in the field of solar power generation. However, they are often affected by defects in photovoltaic panels, which directly harm their performance. In order to solve this problem, this study proposes an improved Mask R-CNN-Attention model for defect detection of photovoltaic cell components. By introducing attention mechanism, the model can focus more accurately on key image areas, and adopts a multi-scale feature pyramid optimization to improve recognition accuracy. The experiment proves that this method has achieved significant improvement in the defect detection of photovoltaic cell components, providing an effective technical means for solving this problem. This innovative method is expected to improve the performance and maintenance efficiency of photovoltaic cell components, bring positive impact to the solar power generation industry, and provide key support for the sustainable performance and maintenance management of photovoltaic cell components.

Aiming at the problem of large current stress of dual active bridge DC-DC converter, a parallel transformerless dual active bridge converter is proposed based on dual active bridge converter, which can reduce the current stress of the converter and improve the transmission power while reducing the cost of the converter. Firstly, the mathematical model of current stress versus transmission power and the soft-switching range of the converter under extended phase shifting are analyzed, and secondly, the current stress of the converter is optimized by using the KKT conditional method by taking the current stress as the objective function and the most constraints in terms of the transmission power and the ZVS characteristics. The superiority of the proposed parallel transformerless dual active bridge converter in terms of current stress is concluded by comparing the losses of the conventional shunt DAB converter with those of the series-parallel type dual active bridge converter. Finally, simulation experiments are carried out to verify the correctness of the theoretical analysis and that the proposed converter circuit is capable of achieving lower current stress and higher power density.

With the continuous advancement of technology, infrared detection technology has become increasingly important in various fields, and the demand for high-performance infrared detectors has also grown. VO₂ thin films have attracted much attention due to their significant changes in resistance and transmittance during phase transitions, especially their high sensitivity to infrared light irradiation at room temperature. In this study, vanadium dioxide (VO₂) thin films were successfully grown on single-crystal sapphire (Al₂O₃) substrates using the hydrothermal method, and the optoelectronic performance of the films was optimized by laser annealing technology. Experimental results show that by adjusting the power of the laser annealing, VO₂ thin films with a good response to 1064 nm infrared light can be obtained. Under a light intensity of 20%, the photocurrent of the film can reach 61.4 μA, which is an increase of 36.4 μA compared to before annealing. The film also exhibits periodic current changes under periodic infrared light irradiation, demonstrating its structural stability and optoelectronic cycling performance. Under the same infrared light irradiation, a larger photocurrent can be generated. This study provides a more cost-effective and sensitive approach for intelligent manufacturing fields such as night-time intelligent driving and the application of intelligent infrared focal plane.

As a crucial component in the industrial field, the presence of defects in Printed Circuit Boards (PCBs) directly affects the quality and compliance of products. Therefore, defect detection in PCBs holds significant research significance. To address the issues of low efficiency, limited accuracy, and poor flexibility in traditional PCB defect detection methods, this paper combines deep learning algorithms. We improve the original YOLOv3 algorithm to detect PCB defects, enhancing the detection accuracy of the network model. Channel pruning, model fine-tuning, and quantization are employed to achieve a lighter network. This paper presents a collaborative approach utilizing convolutional neural networks and FPGA hardware platforms to accomplish PCB defect detection.

Driving distraction is one of the primary causes of frequent traffic accidents. Image data-based driving distraction recognition technology has advantages such as low cost, easy deployment, high efficiency, and accuracy, which is of significant importance for improving traffic safety levels. In this paper, the VGG16 network model is adopted, and the SF dataset is preprocessed, followed by training and testing the network. The experimental results demonstrate that VGG16 achieves an average accuracy of 97.82% in recognizing 10 driving behaviors. Additionally, a circuit design scheme for the various modules of VGG16 based on memristors is proposed, and simulations are performed on the MNSIM platform. Compared to the detection results running on a GPU, the accuracy loss is only around 5%.

In order to study the ultrasonic elliptical vibration-assisted machining device and realize the vibration-assisted machining process, it is of great significance to design a three-dimensional ultrasonic vibration platform for studying the influence of vibration parameters on machining effect and improving machining efficiency. By analyzing several common flexible hinges, a new type of spatial flexible mechanism is designed, and a three-dimensional ultrasonic elliptical vibration platform is designed, and its simulation analysis and test are carried out. The results verify the advantages of the designed flexible mechanism and the effectiveness and correctness of the elliptical vibration platform.

This article establishes a finite element model of a low-temperature energy storage sealed U-shaped spring, and uses the control variable method to analyze the relationship between spring performance and the main influencing factors. The results show that the spring compression reaction force and spring stiffness coefficient increase with the increase of material elastic modulus, spring thickness, and spring radius, and decrease with the increase of spring clearance and cross-sectional arc. Changing the elastic modulus is the simplest way to obtain the required spring force, and the change in spring thickness has the most significant effect on changing the elasticity of U-shaped springs.

This article designs and implements a handkerchief paper vending machine based on air780e microprocessor, aiming at providing an automatic and intelligent public health solution. It adopts the Internet of Things architecture and relies on the existing handkerchief paper vending machine to design. In the sensing control layer, the clever use of micro-switches and mechanical structures, real-time monitoring of electricity, inventory and shipment status, effectively avoid the impact of product shortages, equipment failures on users. In the network communication layer, the vending machine communicates with the server through the MQTT protocol, uses Ali Cloud server to deploy EMQX as the server, and uses 4G network to enhance the data transmission function of the device to ensure that the information interaction between the device and the cloud server is more stable and timely. At the application management level, scan code payment is introduced through the docking payment platform, and a simple cloud-based management system is built, enabling administrators to remotely monitor the operating status and visually reflect real-time sales data. Through the application of Internet of Things technology, this design provides reference experience for related industries and shows the great potential of modern science and technology in public health services.

As the global population ages, the design of smart home products for the elderly plays a crucial role in improving their quality of life. To enhance elderly satisfaction with these products and help designers understand their needs accurately, we propose a design method based on Structural Equation Modeling (SEM). We first collect requirements from the elderly and designers through interviews and questionnaires, transforming them into design elements. Using factor analysis, we categorize these elements. Then, with SEM, we analyze the priority of design elements quantitatively and propose relevant design strategies. The paper establishes an innovative design process guided by user needs, providing a systematic approach and strategies for the design of smart home products for the elderly.

In this paper, a high-precision temperature measurement method based on thin-film thermocouples is proposed for the key demand of intelligent transformation of the construction industry, which focuses on improving the ability to perceive the construction environment, especially the temperature monitoring of workers' helmets. A good temperature measurement system is constructed experimentally. Experimental oven was used as the heat source, room temperature as the cold source, and a digital multimeter and an infrared thermometer were used to ensure that the testing conditions were controllable and the data acquisition was accurate. The data were carefully analysed using the least squares method by heating the thin film thermocouple and recording the electric potential at different temperatures. The results show that the adopted quadratic polynomial fitting model possesses a high degree of linear correlation (R² >0.999), indicating a strong linear relationship between the thermoelectric potential and temperature. The average Seebeck coefficient obtained from experiments was 41.46 μV/ °C, which is highly compatible with type K thermocouples, verifying the validity and practicality of the calibration method. In addition, the high repeatability and stability of the temperature measurement system was confirmed by comparing the data from several experiments. This study not only provides an important experimental basis and technical support for the environmental sensing technology in the field of intelligent construction, but also lays a solid foundation for promoting the process of construction automation and intelligence, which has significant theoretical significance and application value.

Due to the problem of low control accuracy in traditional robotic arm movement path control methods, a research on bionic robotic arm movement path control method for non climbing operations is proposed. This method first determines the structural parameters and kinematic characteristics of the bionic robotic arm through parameter measurement, providing basic data for subsequent control method design. Secondly, a kinematic model of the biomimetic robotic arm was constructed, and structural and kinematic analyses were conducted, providing a reliable simulation platform for path planning and control. Finally, utilizing inverse kinematics control, intelligent and precise motion control of the robotic arm was achieved in highly hazardous environments. The experimental results show that the control accuracy and efficiency of the method proposed in this paper are higher, and the control effect is better.

In order to adapt to the high-pressure environment generated by the phase change of carbon dioxide and the characteristics of rapid reaction, a pressure-controllable quick-opening valve device was designed. The device uses the pressure ratio between the large and small cylinders to control the piston movement, and when the working pressure ratio exceeds the calculated pressure ratio, it will push the piston to move to complete the valve opening action and release the pressure. Through the establishment of the valve model, calculate the pressure ratio between large and small cylinders, analyse and simulate the process of opening the valve, derive the opening time and the pressure change formula of large cylinders and carry out simulation calculations. The principle prototype test was carried out, and the test results show that the valve can meet the requirements of opening pressure control and millisecond rapid response. This not only provides a new model for supercritical carbon dioxide launchers, but also has a wide range of application prospects in related fields under similar working conditions such as liquid gas transmission.

Ultrasonic residual stress detection has many highlights including but not limited to fast detection speed, no radiation damage, good spatial resolution, and deep detection range. The detection accuracy of residual stress using the critical refractive longitudinal wave method is affected by several factors. The study of these effecting factors is helpful to improve the accuracy of ultrasonic nondestructive testing of residual stress. In this paper, the measurement errors caused by temperature and coupling state are investigated. The results show that for the one-transmitter & one-receiver (OTOR) probe, changes in the coupling state and temperature can cause large deviations. The use of one-transmitter & doublereceiver (OTDR) probe can effectively reduce the error caused by temperature and coupling state, and improve the accuracy of the detection system.

This paper introduces a method for detecting tunnel water leakage diseases by combining the YOLOv8 network with a Bayesian reasoning model. Initially, the YOLOv8 network is utilized to construct a dual detection model for identifying the tunnel water leakage diseases and leakage-prone area. Subsequently, a tunnel water leakage disease screening model was constructed based on Bayesian reasoning. By employing the probability weight of the leakage-prone area, the confidence of the tunnel water leakage disease, and the probability weight of the prone area have tunnel water leakage diseases are combined to ultimately obtain the probability of tunnel water leakage diseases, this probability will be used to reduce false alarms in the detection results. Experimental testing using actual tunnel data demonstrates that compared to the YOLOv5 and YOLOv8_m networks, the proposed model achieves an increase in mAP by 7.37% and 1.64% respectively, effectively validating the model's efficacy.

The running data of multi-power watt-hour meters is large, which is easy to produce relative errors, and the monitoring results are missed, and the monitoring effect is not ideal. Therefore, a centralized remote monitoring method of multipower watt-hour meters based on Internet of Things technology is proposed. A multi-parameter monitoring framework of Internet of Things for multi-power watt-hour meters is constructed. Based on this framework, a centralized remote monitoring model of multi-power watt-hour meters is established, and the multi-parameter weights of multi-power watt-hour meters are adjusted according to the aging coefficient of multi-power watt-hour meters. The filter function is used to adjust the multi-parameter weights of multi-power watt-hour meters to determine the running state of multi-power watt-hour meters and realize centralized remote monitoring of multi-power watt-hour meters. The experimental results show that the relative error and missed detection rate of this method are small, and the monitoring accuracy and stability are excellent, which provides more accurate and effective technical support for intelligent management of power system.

In order to compare the performance of time-differentiated fluxgate sensors under different conditions and to study the specific waveform selection problem, a simulation modelling method for constructing time-differentiated fluxgate sensors based on Matlab/Simulink software is proposed. The simulation results show that the established simulation model is convenient, efficient and operable, and the simulation output can effectively and accurately reflect the effects of different waveforms on the sensor performance. The model proposed in this paper can not only simulate the measurement process of the time-differential fluxgate sensor, but also provide a model basis for the study of the performance of this sensor. Indepth analysis and study of the model can let us understand the working principle and performance characteristics of the time-difference fluxgate sensor more clearly, so as to provide help for the optimisation of fluxgate sensor parameters and performance improvement.

The asymmetric flow between facilities and the location of fixed facilities are ignored in the current unidirectional loop layout research, which is inconsistent with the actual situation of the production shop. On this basis, the unidirectional loop layout problem considering the asymmetric flow between facilities and the specific location of facilities is proposed. To minimize the material handling cost, a mixed integer programming model is constructed, and the precision solution optimizer LINGO is used to solve small-scale examples to verify the correctness of the model. In order to improve the efficiency of the algorithm, heuristic rules are designed to ensure that all the solutions are feasible when generating the initial solutions, and the genetic algorithm is improved by combining adaptive crossover, adaptive mutation and nested simulated annealing algorithm. Finally, the double threshold stop criterion is used to remove the number of redundant cycles. Firstly, the algorithm is used to solve small-scale problems, and the results are compared with the exact solutions to verify the correctness of the algorithm. Secondly, the proposed algorithm is used to solve 15 examples in the 4-49 scale, and the results are compared to verify the efficiency of the proposed algorithm. Finally, taking the field investigation enterprise as an example, the layout of the security door production line in the production workshop was optimized, and the proposed adaptive genetic simulated annealing algorithm was applied to solve the problem.