To accurately predict the standard man-hours of aircraft maintenance tasks, a predictive method based on safe energy consumption for aircraft maintenance standard man-hours was proposed. Firstly, a drivable virtual human was constructed based on the skeletal parameters of the human body and the computational rules of Euler quaternions. Subsequently, motion data during maintenance processes were captured using motion capture devices. The Roberson-Wittenburg and inverse dynamics methods were then employed to calculate the joint torques borne by the virtual human during the maintenance tasks, thereby estimating the energy consumption during the maintenance process. Based on the fatigue assessment of human energy consumption, preliminary standard maintenance man-hours conforming to safety energy consumption were calculated. The study demonstrates that prediction results based on safe energy consumption are closer to the manufacturer's recommended times, with a smaller root mean square error than modular arrangement of predetermined time standards(MOD) method, indicating that this approach better reflects the actual maintenance man-hours.

In order to reduce the occurrence of building construction safety accidents, association rules were used to reveal the mechanism of accident association, and the optimized RF was fused to predict the occurrence of accidents. First, the causal factors of 388 case reports of construction safety accidents were extracted using 24Model as the theoretical basis. Then, Apriori algorithm was used to excavate the interrelated action paths between the accident causal factors. Finally, hyper-parameters of RF were optimized using GWO algorithm, and the GWO-RF prediction model of construction safety accidents was constructed. And the accident causal factors were the characteristic importance ranking was carried out. The results show that: unsafe behavior, safety ability of organization members, safety management system and safety culture elements constitute a combination of strong correlation conditions. GWO can effectively optimize the hyper-parameters of RF, and prediction accuracy of the optimized GWO-RF model is as high as 93.2%. The characteristic importance ranking shows that: safety education and training have the greatest influence on the prediction of construction safety accidents, with a weighting of 10.5% and a weighting of 10.5%. The importance ranking of features shows that: safety education and training has the greatest influence on the prediction of building construction safety accidents, with a weight of 10.5%. And safety integration management, safety production rules and regulations, and safety production responsibility system are the important factors affecting the prediction of building construction safety accidents, with weights of 7.5%, 7%, and 6%, in that order.

To ensure the normal operation of the airport terminal system, the impact of redundancy on operation resilience was analyzed by taking incidents such as security checkpoint failures and staff absenteeism at check-in counters as examples. A definition and quantification method for resilience, covering the entire process from system disruption to recovery, was proposed. T1 terminal of a major airport handling tens of millions of passengers annually was selected as a case study. An operation model of the terminal was constructed based on Anylogic. The influence of the number of security checkpoint failures, failure duration, recovery measure intensity, and the number of absent staff at manual check-in counters on terminal operation resilience was analyzed under scenarios with and without redundancy. The results indicate that when the system redundancy is lacking, the operation resilience of the terminal decreases as the number of security checkpoint failures and absent staff increases, shows no significant correlation with failure duration, but improves with the enhancement of recovery measure intensity. In contrast, when the system incorporates redundancy, the functional level either remains unaffected or declines less significantly compared to the non-redundant scenario, and the system resilience is higher. Moreover, if the system possesses redundancy and the redundancy degree exceeds the disruption intensity, the functional level remains unchanged, and the system resilience approaches 1.0. Therefore, redundancy is a critical factor that cannot be overlooked in resilience analysis.

To improve the efficiency of UAV air traffic management, ensure flight safety and promote the safe application of UAVs in complex airspace environments, the risk assessment of UAV operation was focused in this paper. Firstly, for UAVs with autonomous perception and decision-making capabilities in unstructured airspace environments, based on key parameters such as airborne communication, navigation and surveillance capabilities, maneuvering characteristics and system response time. A conflict probability model and a collision probability model considering avoidance maneuvering strategies were constructed to quantitatively evaluate the airspace collision risks. Then, considering that UAV collision accidents do not directly cause casualties, a ground risk assessment model was constructed that comprehensively considers UAV aerial collision events and crashes caused by system failures. Finally, taking 1×10^-6 deaths/flight hours as the safety target level, the safety separation that needs to be maintained during airborne flights was determined. The results show that considering both the conflict probability and the probability of the conflict escalating into a collision simultaneously can solve the problem of underestimated risks during the free flight stage. The maximum allowable collision risk varies significantly among different operating scenarios.

To improve the accuracy and real-time performance of identifying unsafe behaviors of miners in the mine belt transportation area, and to address the problems of poor real-time performance and high false detection rate in existing manual monitoring methods, a dual-stream spatiotemporal fusion network (DS-SFNet) that integrated image features and human skeleton features was proposed. First, challenges such as low illumination and dust interference in underground environments were addressed by designing a sub-pixel convolutional block attention module (SPCBAM), which combined with sub-pixel convolution and depth wise separable convolution to optimize feature representation. Second, to mitigate the high computational resource consumption of the OpenPose model, its backbone feature extraction network was reconstructed using MobileNet v3 by incorporating dilated convolutions and cross-layer connections. Finally, a hierarchical feature fusion module was constructed to deeply integrate image features and skeletal trajectory features through spatiotemporal alignment and complementary modeling. The results demonstrate a recognition accuracy of 76.4% on HMDB51 (Human Motion Database 51) and 97.9% on UCF101 (University of Central Florida 101), outperforming the SlowFast model by 1.5% and 1.1%, respectively. On a self-built coal mine dataset containing four unsafe behaviors (climbing, crossing, leaning, and hand-leaning), the average recognition accuracy reaches 92.3%. The MobileNet v3-reconstructed OpenPose model reduces parameters to 11.5% of the original Visual Geometry Group 19 (VGG19) network while increasing inference speed by over 3 times. The complete framework achieves a single-frame processing time of 38.7 ms and a parameter count of 57.3 M.

In order to enhance the intelligent diagnosis level of safety problems in complex construction environments, an intelligent question-answering model for construction safety hazards based on vision-language multimodality was proposed. A dataset of image-text pairs related to construction safety hazards was constructed. A visual encoder was used to complete the visual encoding of safety hazard images, and a language model was employed to encode the question-answering texts about safety hazards. A multimodal feature fusion module was adopted to achieve effective interaction between image and text information. A specific input template for visual question answering adapted to the scenario of construction safety hazards was constructed. The model was fine-tuned based on matrix low-rank decomposition, and multi-round prompts were used to guide the model in generating accurate answers. The results show that compared with existing contrastive models, the intelligent question-answering model for construction safety hazards performs better in automatic evaluation metrics, Generative Pre-trained Transformer(GPT)-4 evaluation, and expert evaluation, with significantly improved fluency and semantic relevance of the generated texts. Ablation experiments further verify the effectiveness of each sub-module, confirming that the synergistic effect of matrix low-rank decomposition fine-tuning and multi-round reasoning is the key for the model to achieve optimal performance, and that reasonably setting the rank parameter of the low-rank matrix can effectively avoid the overfitting problem.

To address the issues of insufficient resolution in the high-density electrical resistivity method and limited accuracy in anomaly indentification for road collapse hazard detection, resolution tests for road collapse hazard detection based on high-density electrical resistivity method and investigation of an anomaly identification method using GMM clustering were conducted. Forward modeling was performed using the finite difference method, while inversion process was carried out using the Gauss-Newton method. Numerical simulations were conducted to assess the effect of different electrode spacing configurations on detection resolution. In the context of pipeline leakage-induced road collapse, geoelectric models for underground anomalies at various stages of development were designed, and GMM clustering analysis was applied to optimize the inversion results of the high-density electrical resistivity method. The results demonstrate that adjusting the electrode spacing and measurement parameters can significantly improve detection resolution. At a depth of 4.5 meters, the location and shape of underground anomalies at a scale of 1 meter can be effectively characterized by reducing the electrode spacing. An electrode spacing of 0.5 meters can balance detection accuracy and computational efficiency, corresponding to half the scale of the target anomaly. For anomalies buried at the same depth, the resistivity recovery of low-resistance anomalies is superior to that of high-resistance anomalies, providing the basis for parameter optimization for detecting various anomaly types. The feasibility of high-density electrical resistivity method to detect leakage-induced detects at different stages is demonstrated through tests on underground cavity models induced by pipeline leakage, while the identification accuracy of anomaly regions is further enhanced by the GMM-based clustering analysis.

In order to enhance driving safety in highway tunnels and to reduce visual distraction and burden caused by improperly designed visual guiding devices, a research framework for the design and evaluation of tunnel visual guiding devices was proposed. A systematic review of domestic and international literature has been conducted to analyze the effectiveness and design parameters of various types of visual guiding devices, and to clarify their key functions in delineating spatial right-of-way, ensuring adequate sight distance, and enhancing visual comfort. In addition, current studies were reviewed to summarize data acquisition methods and core evaluation indicators, thereby constructing a scientific evaluation system. The limitations of existing research were also identified, and future research directions were proposed. The results show that visual guiding devices can optimize the tunnel visual environment by clarifying spatial right-of-way, increasing effective sight distance and visual fields, and mitigating visual illusions. The type and spacing of devices should be determined based on spatial right-of-way, curvature perception, speed perception, and visual comfort. Real-vehicle experiments and indoor simulations serve as key approaches for exploring design parameters and optimization strategies, while field observations and accident data analysis can further validate their effectiveness. Indicators such as visual perception, visual characteristics, physiological responses, and driving behavior are suggested as core evaluation metrics. Moreover, multi-indicator interaction analysis may help reveal the mechanisms by which visual guiding devices influence driver behavior.

To achieve dynamic risk management and control during the CO₂ injection process in CCUS oilfields and prevent CO₂ leakage acidents, a dynamic assessment method for the probability of high-risk scenarios was proposed. Firstly, a novel improved FPN model was developed by integrating a three-parameter Weibull distribution model for equipment failure probability based on Grey Model (GM) and Support Vector Machine (SVM) with FPN model. Subsequently, the risk analysis was conducted on the above ground CO₂ injection system of CCUS process. The CO₂ leakage accident chain was established to quantitatively predict the accident occurrence probability. The dynamic characteristic of equipment failure probability changing with time (t) and the impact of protection layers on risk were taken into account by the improved FPN. The results indicate that when t=500 h, the risk exhibits an exponential increase. The safety valve is identified as the protection layer with the highest importance. When t=303 days, the system's residual risk reaches the PetroChina risk acceptance criterion of 1×10^-5, indicating that the system risk becomes unacceptable beyond this point. It is recommended to add a high-high level interlock protection layer, which extends the time until system risk becomes unacceptable to 5 832 days, thereby significantly reducing maintenance frequency and costs.

In order to prevent structural safety accidents in commercial buildings during the operation and maintenance phase and enhance the objectivity and accuracy of structural safety assessments for commercial buildings, the following steps were taken: First, a structural safety evaluation system for the operation and maintenance phase of commercial buildings was established. The weights of each evaluation indicator within the system were determined using a combined objective-subjective weighting method based on the Analytic Hierarchy Process (AHP) and the Criteria Importance Through Intercriteria Correlation (CRITIC) method. Second, by incorporating the improved TOPSIS method with interval numbers, the interval values of safety evaluation indicators corresponding to different structural safety levels of commercial buildings were determined, yielding relative proximity values under different safety condition levels. Finally, taking a commercial complex building in Changsha City as an example, the safety condition was evaluated based on on-site inspection data, and the results were compared with those from the inspection report issued by a professional inspection institution after appraisal to verify the accuracy and applicability of the established indicator system and calculation results. The study indicates that by combining industry standards and expert experience to establish the evaluation indicator system, and using the combined weighting method and the interval-based TOPSIS evaluation method improved with interval numbers on the basis of existing research, the objectivity and accuracy of structural safety assessment for commercial buildings can be enhanced.