In the intersection scenario, the running state of social vehicles, the control state of traffic lights and the accurate identification of track components have become the technical bottlenecks restricting the promotion and application of track inspection robots. Aiming at the requirements of track health inspection, firstly, the vision inspection system of the track inspection robot and the technical scheme of the navigation system was presented based on “Beidou +5G”. Secondly, the vision detection system model was built based on YOLOv8 algorithm, and the web crawler technology was innovatively used to capture sample data about traffic lights and car taillights from open source video resources to train the vision detection model. Then, transfer learning method and early stop method were used to optimize the detection accuracy of the trained model. The research results show that after adopting YOLOv8 algorithm and optimizing the model with transfer learning method and early stop method, the inspection robot can effectively detect the track components, vehicles and traffic lights at the switch junction, and effectively improve the inspection efficiency and accuracy.

To solve the problem that the wire rope swing angle is too large during the load lifting and lowering process of the bridge crane, a fuzzy layered sliding mode control method was proposed with the single pendulum model of the bridge crane as the research object. The method firstly establishes the bridge crane single pendulum model system, designs the two-layer sliding mode surface joint control based on traditional sliding mode control, combines the layered sliding mode surface with fuzzy control to create the controller, proves the Lyapunov stability of the closed-loop system of the bridge crane under the method through theory and carries out simulation experiments. The simulation results show that compared with the LQR controller and the multi-sliding mode controller, the trolley arrives at the desired position in 68% less time, and the maximum swing angle of the load is reduced by 15%, which achieves a good anti-swing effect.

Aiming at the problems of nonlinearity, large delay, uncertain model parameters and weak anti-interference ability in the temperature control system of variable air volume air conditioning supply air, a closed-loop control system of supply air temperature based on active disturbance rejection control (ADRC) was designed. In order to overcome the difficulty in parameter adjustment of active disturbance rejection controller, an improved grey wolf optimization (IGWO) algorithm was proposed to optimize controller parameters. By introducing chaotic mapping, nonlinear convergence factors, dynamic weights and dimensional learning strategies into grey wolf optimization (GWO), the population diversity was increased and the balance between search and exploitation was taken into account. The advantages and feasibility of the proposed algorithm were verified by MATLAB simulation. The experiment further proves that compared with the traditional proportional integral derivative (PID) controller and the traditional gray wolf algorithm, the IGWO algorithm can shorten the supply air temperature overshoot by 45.3% and 8.9%. The adjustment time is reduced by 34.8% and 11.2%, the steady-state error is smaller, and the system is more energy efficient.

The Unmanned aerial vehicles three-dimensional path planning problem is a combinatorial optimization problem to find the optimal path between the starting point and the endpoint in complex three-dimensional environment, but most path planning algorithms struggle to find feasible paths within acceptable time and precision range, therefore, a dynamic multi-subswarm salp swarm algorithm based on K-means++ clustering optimization was proposed to address the aforementioned issue. Firstly, a new cost function incorporating height cost was proposed within the three-dimensional environment model. The path planning problem was converted into a multi-dimensional function optimization issue. Secondly, the population was clustered using the K-means++ clustering algorithm, and a dynamic multi-subswarm mechanism was designed to balance the algorithm's global search and local exploitation. Each subswarm collaborates with multiple strategies for improvement, avoiding the algorithm from being trapped in local optima while enhancing global optimization capability. Finally, after validating the algorithm against five algorithms ISSA, MSNSSA, IBSO, MBFPA, and SSA using 12 CEC2017 benchmark test functions, it was applied to solve the optimal path planning problem in three-dimensional environments. Simulation results under different environmental models demonstrate that the algorithm's average effective path rate is increased by 15.5%, 11%, 23%, 20.5% and 18% compared to the other five algorithms, confirming its excellent optimization capability in complex environments.

Existing voice-driven facial generation methods still face challenges in feature extraction and generation quality, and have yet to fully explore the deep correlation between audio and facial features. To address above mentioned issues, a research approach that combines Mel frequency cepstral coefficients (MFCC) was proposedfor audio feature extraction with the image generation capabilities of the second generation of style generative adversarial networks (StyleGAN2) was proposed. In terms of audio processing, MFCC was employed as the feature extraction method. To more effectively extract and transmit features from the audio, a ResNet18-based residual module was designed and integrated with the squeeze-and-excitation (SE) attention mechanism. Additionally, the activation function in the original residual blocks was optimized and improved by using the Mish activation function, aiming to mitigate the gradient vanishing problem in deep networks, maintain the integrity of feature information, and enhance the accuracy and generalization ability of the model. The StyleGAN2 model was then utilized as the facial image generation model. Experimental results demonstrate that the integration of the designed audio processing network with the StyleGAN2 facial generation model exhibits outstanding performance in the task of voice-driven facial generation. Through comprehensive evaluation using metrics such as Fréchet inception distance (FID) and path length, the proposed method shows a significant improvement in generation quality compared to existing methods, thus fully proving its effectiveness and superiority.

An algorithm has been proposed to detect small targets in unmanned aerial vehicle(UAV) aerial images. The algorithm is based on an improved real-time detection Transformer (RT-DETR) and aims to address the challenges posed by complex backgrounds and a large number of small target samples. To enhance the feature fusion network, a dedicated feature fusion structure for small targets has been incorporated, utilizing rich location information from the shallow feature map to improve the network's ability to detect small targets. Furthermore, the last residual block in the BackBone has been removed to prevent an increase in additional parameters. Additionally, the MCP Block, a reconstructed BasicBlock structure in the backbone network, has been designed, which includes a multi-channel feature partial convolution module (MCPConv) to reduce redundancy in channel features and enhance the acquisition of multi-scale detail features. Moreover, a location encoding mechanism with learning ability has been introduced to obtain more accurate and expressive location information. The normalized weighted deviation(NWD) and mean precision-driven IoU(MPDIoU) positioning loss functions have been incorporated to accelerate the convergence speed of the model and reduce sensitivity to position deviation. Experimental results on the VisDrone2019-DET dataset demonstrate that the improved model reduces parameters by 62% compared to the original model, increases mAP50 by 3.9%, and improves FPS by 17%. The improved model exhibits superior detection performance compared to other mainstream detection models.

Aiming to address the issue of accurately measuring the slurry pH value during the operation of limestone-gypsum wet flue gas desulfurization (WFGD) system, which hinders the efficient operation of WFGD, a prediction model for FGD system pH based on bi-directional gated recirculation unit(BiGRU) has been developed. Firstly, the raw data were cleaned and normalized. Secondly, based on the maximum information coefficient analysis, 13 characteristic values were obtained as input variables and pH as output variables, and a slurry pH model was established. Finally, the model was run and the results were evaluated. Compared to LSTM and GRU, the results indicate that the mean absolute error of this mathematical model decreases by 11.95% and 24.92%, while the root mean square error decreases by 10.64% and 19.49%. Additionally, the coefficient of determination improves by 1.79% and 3.08% respectively. This demonstrates that the BiGRU-based pH predictive model exhibits high accuracy and stability, making it valuable for engineering applications and providing an important reference for predicting pH models in existing desulfurization tower systems.

Molecular dynamics method is adopted to investigate the effect of different NaCl solutions concentrations on the bonding properties of the calcium silicate hydrate/γ-FeOOH(C-S-H/γ-FeOOH) interface. The effect mechanism of NaCl solution concentration is revealed from the interface ion evolution, radial distribution function, particle strength distribution, interaction energy and mechanical properties. The results show that as the concentration of NaCl solution increases, interlayer {\mathrm{Ca}}_{\mathrm{h}}^{2+} ions separate from the surface of C-S-H and diffuse to the interlayer solution, Na⁺ ions enter the C-S-H layer. {\mathrm{Ca}}_{\mathrm{h}}^{2+} ions adsorb Cl^- ions in the solution, resulting in the ion clusters of {\mathrm{Ca}}_{\mathrm{h}}^{2+} and Cl^- on the surface of C-S-H. In addition, the γ-FeOOH surface hydroxyl oscillation provides adsorption points for {{\mathrm{Na}}^{+}}^{}ions, resulting in the increase of Na⁺ ions on the γ-FeOOH surface. When the NaCl solution concentration increased, the RDF peak of Ca_h—O_s gradually decreased and the radial distribution function(RDF) peak of Ca_h—O_w, Ca_h—Cl, and Na—O_s gradually increases, consistent with the ionic strength distribution. Where, O_s is the oxygen on the silicon chain in C-S-H, and O_w is the oxygen in the interlayer solution water. {\mathrm{Ca}}_{\mathrm{h}}^{2+} ions form ionic bonds with O_w in water, leading to a reduction of Ca_h—O_s ionic bonds on the C-S-H surface. Since the strength and stability of Ca_h—O_s ionic bond are better than that of Ca_h—O_w,therefore, the C-S-H/γ-FeOOH interfacial interaction energy and peak stress both show a decreasing trend with the increase of NaCl solution concentration.

In view of the deficiency of the research on the prevention and control of bolt anchoring in engineering rock mass, the method of cooperative prevention and control analysis based on bolt pre-tightening force was put forward, and the cooperative prevention and control test of bolt, anchor agent and surrounding rock based on bolt pre-tightening force was carried out, the variation law of pre-tightening force of bolt was obtained and the cooperative prevention and control state of bolt anchoring was identified. The results indicate that improving the synergy between anchoring agents and the surrounding rock of anchoring holes, as well as between anchor rods and anchoring agents, is beneficial for their synergistic evolution and can effectively enhance the prevention and control effect. A method for determining the collaborative state of anchor rod anchoring is provided, which can comprehensively determine the collaborative state of anchor rod anchoring through the relaxation process curve and relaxation degree of anchor rod pre-tightening force. Increasing the contact surface between the anchor rod pad and the surrounding rock on the free face is beneficial for their synergistic evolution. Based on the application and monitoring of pre-tightening force, a method for determining the overall coordination degree of anchor rod anchoring prevention and control has been developed, and strategies for improving the coordination degree of each part of anchor rod anchoring have been proposed. During design, special attention should be paid to the coordination of each part of anchor rod anchoring to ensure that they can perform at their best and be in their optimal state. The research results have good guidance and reference significance for the anchoring mechanism, monitoring, prediction and prevention of prestressed anchor rods in engineering rock masses.

“Large-difference annulus” is a common characteristic encountered in complex wellbore structures with “varying diameters” during (ultra-)deep well drilling. This characteristic leads to slow drilling fluid velocity in the upper large-diameter annulus, posing challenges for cuttings removal, while the higher velocity in the lower small-diameter annulus results in significant circulating pressure loss. To address these issues, a novel flow diverter tool was designed to carry cuttings in the upper section and the loss of circulating pressure in the lower section. However, the lack of a specific wellbore flow model tailored for this tool in current research has hindered its design and optimization. Based on the fundamental principles of fluid flow and hear transfer, a valid mathematical model was proposed to be compatible with the flow diverter tool. Then, via a case study on Well ZS102, the tool was proved to be effective and perform well in practice. The results show that with the installation of the flow diverter tool, the bottomhole pressure is lowered from 85.08 MPa to 80.30 MPa, the standpipe pressure is significantly reduced from 20.97 MPa to 7.22 MPa, and the annulus pressure loss is decreased from 7.16 MPa to 2.40 MPa. The research presents a novel approach for optimizing cuttings removal parameters and preventing leaks in complex wellbore structures during deep and ultra-deep well drilling operations, contributing significantly to the advancement of related drilling technologies.