To ensure the safety of personnel and property within the storage environment，the traditional YOLOv11 object detection algorithm was improved，and a method and model to identify unsafe behaviors of personnel in the complex environment of tobacco warehouses were proposed. First，a statistical analysis of common unsafe behavior types in tobacco storage was conducted，and the classification of unsafe behaviors of warehouse personnel was explored，including item-related，action-related，and area-related unsafe behaviors. Second，based on the characteristics of unsafe behaviors of warehouse personnel，a dataset augmentation and denoising preprocessing approach was proposed to enhance fine-grained feature extraction，and introduced to improve the saliency mapping of personnel behaviors. Then，the YOLOv11 algorithm was improved through functional enhancement modules and K-means++ anchor box optimization，and a fast detection method for unsafe behaviors of tobacco warehouse personnel was proposed. Finally，the proposed method's effectiveness was validated by comparing with self-built datasets and the open Microsoft COCO dataset. The results show that the method can quickly and effectively identify unsafe behaviors of warehouse personnel，with a significant improvement in recognition accuracy compared to traditional methods(accuracy rate is 94.91% and 88.69% respectively).

To address the safety risks caused by rapid fire spread and thermal smoke accumulation in confined spaces such as basements and underground warehouses due to lithium-ion electric bicycle fires，this study combines experimental and numerical simulation methods to investigate the temperature field evolution，flame characteristics，and smoke diffusion behavior of lithium-ion electric bicycle combustion under different ambient temperatures，clarifying the effect of temperature on combustion characteristics. The results show that under experimental conditions，the average temperature at the combustion center in the confined space is approximately 600 ℃，with a peak temperature exceeding 920 ℃. Ambient temperature significantly affects the initial stage of lithium-ion electric bicycle combustion. In a 40 ℃ environment，the time for combustion to enter the rapid development phase is shortened by 20 seconds compared to 20 and 0 ℃ environments. The time required for nearby temperatures to reach the ignition point of the lithium-ion electric bicycle is reduced by approximately 25 seconds compared to 20 and 0 ℃ environments. After about 80 seconds，the temperature rise rates converge. At 20 seconds，the flame morphology of lithium-ion electric bicycle combustion differs noticeably across environments：the flame height at 40 ℃ is approximately 1.15 times that at 20 ℃ and 1.32 times that at 0 ℃. Flame morphology converges after about 80 seconds. Within the first 30 seconds，the smoke diffusion velocity and production rate in a 40 ℃ environment are significantly higher than those at 20 and 0 ℃，but smoke concentrations stabilize to similar levels after approximately 50 seconds.

In order to further study the detonating effect of UAV，and improve the fire extinguishing efficiency of fire bombs，ANSYS Workbench software was used to establish a finite element model of fire bombs，and the explosion and dispersion process of fire extinguishing agent was simulated to clarify the impact of different detonation heights on the dispersion characteristics of fire extinguishing agent. Meanwhile，full-scale experiments on the vertical dropping of fire extinguishing bombs by unmanned aerial vehicles at different detonation heights were conducted. The results show that the simulation data such as the explosion process of fire extinguishing bomb，the throwing process of fire extinguishing agent and the spreading radius of fire extinguishing agent are in good agreement with the full-scale test results. After the fire extinguishing bomb explodes，the extinguishing agent disperses in the air in a cone shape and spread evenly in the horizontal direction，and the spreading shape is roughly circular. With the increase of detonation height，the landing time of fire extinguishing agent increases，and the horizontal velocity of fire extinguishing agent decreases. The increase of landing time plays a dominant role relative to the decrease of horizontal velocity，resulting in the increase of the spreading radius of fire extinguishing agent with the increase of detonation height. When the detonation height of the fire extinguishing bomb increases from 5 to 12 m，the numerical simulation data of the fire extinguishing agent spread radius increases from 2.04 to 3.56 m，and the error between the numerical simulation data and the experimental data is within 5%.

In order to systematically explore the root causes of industrial hot work accidents through a large amount of text data，a text mining method based on 24Model was proposed. Firstly，220 hot work accident reports were collected and sorted as datasets，and a 24Model classifier based on Bidirectional Encoder Representations from Transformers (BERT) was constructed. The pre-trained model was used to train and evaluate the accident report dataset to construct a classification model. Then，through the combination weight of the Keyword extraction algorithm based on BERT (KeyBERT) and Term Frequency-Inverse Document Frequency (TF-IDF) algorithms，combined with the 24Model framework，a keyword index system for hot work accident text was established. Finally，the interrelationships between accident causes were obtained through the analysis of the network co-occurrence relationship between text-mining keywords. The results show that the BERT-based 24Model classifier model can systematically and accurately determine the causative categories of hot work accidents. The weight of the safety management system was the largest among the 4-level keyword index systems obtained through the combination of weights. Furthermore，7 key causative factors of hot work accidents were obtained by combining them with the co-occurrence network analysis. This shows that 24Model can strengthen the interpretability of text mining results，which provides an important reference for the prevention and management of hot work accidents.

In order to effectively prevent cross operation accidents in the construction process，firstly，130 cross operation accident cases nationwide from 2010 to 2022 were collected as research objects，and an evaluation system for cross operation in construction was established based on the grounded theory. Then，according to the relationship between the indicators，a spectral clustering algorithm was used to analyze the combination of accident causes. The weight of each cause factor was allocated，and different levels of risk were divided by principal component analysis and analytic hierarchy process. Finally，the Apriori algorithm was used to obtain the correlation chain of each factor in the causative combination. The results show that the evaluation system includes 5 core categories，13 main categories and 40 categories. The core category that has the greatest influence on the risk of cross operation in the construction process is technical management. The 40 categories are grouped into 4 causa l combinations，among which 1 is high risk，1 is medium risk，and 2 are low risk. In addition，for the complex cluster of factors，the key combination of causes is identified and targeted preventive measures are proposed to reduce the probability of cross-operation accidents.

To reduce building risks in landslide-prone areas and guide disaster mitigation measures，a comprehensive vulnerability analysis method was proposed based on Entropy Weight Method(EWM)-Analytic Hierarchy Process(AHP) and trapezoidal-semi-trapezoidal membership functions. Using the Sifangbei landslide in Wanzhou District，Chongqing，as a case study. The landslide failure probability was calculated using the Morgenstern-Price and Monte Carlo methods. Field surveys of building attributes were used to analyze landslide hazards and validate the analysis method. The results indicate that the composite model using crack indicators provides reliable vulnerability analysis for buildings in the Sifangbei landslide zone. The proposed method is feasible. The landslide is most dangerous when the reservoir water level drops from 175 m to 145 m. Combined with a 50-year rainfall event，leading to over 7.5 million yuan in economic losses. Vulnerable buildings are mainly located in the middle and rear parts of the landslide and along its right boundary.

To prevent miners from mistakenly entering dangerous areas，a lightweight underground miner object detection model based on YOLOv5s-MPD was proposed，which combined with depth camera to locate miner targets and detect whether miners had entered dangerous areas in real time. Specifically，the MobileNetv3 lightweight neural network was used as the backbone feature extraction network to significantly reduce the model size. Secondly，Polarized Self-Attention (PSA) module was introduced to enhance the perception of targets. Finally，Deformable Convolution Network v2 (DCNv2) was used to replace the standard convolution in the C3 module of the feature fusion layer，solving the problem of partial feature information loss in conventional convolution. The improved model was used in combination with the color images obtained by the depth camera to detect miner targets and obtain the spatial three-dimensional coordinates of the target center points. The results show that compared with YOLOv5s，the improved model reduces the number of parameters and computation by 83.54% and 77.03%，respectively. The model size is only 3.4 MB，and a detection speed of 70.2 f/s，which is increased by 54.97%. The mean average precision is 0.825. Compared with mainstream object detection models，the improved model has a more balanced number of parameters，computation，model size，detection speed，and mean average precision. In the actual positioning accuracy test，within a range of 1-8 meters，the average absolute error and average relative error of the distance between the camera and the miner target were 0.11 meters and 1.74%，respectively. The maximum absolute error and maximum relative error were 0.25 meters and 2.96%，respectively. In the dynamic detection，the miner target could be detected and its location information output，with a detection success rate of 97.5%.

To meet the requirements of the future Trajectory-Based Operation (TBO) mode，improve air traffic safety and efficiency，and obtain executable four-dimensional (4D) trajectory results with multi-point control，this study first established a fundamental 4D trajectory dynamics model by integrating atmospheric environment modeling，aircraft point-mass motion modeling，and performance modeling. Subsequently，a 4D trajectory simulation framework was developed using Simulink to validate the executability of the fundamental dynamics model. Following this，an optimal control-based 4D trajectory waypoint control model was constructed with reference trajectories as optimization targets，where the waypoint control problem was transformed into a nonlinear programming problem through the Radau pseudospectral method. Finally，comparative analyses were conducted using historical flight data from QAR. Results demonstrate that the simulated velocity and mass parameters exhibit zero deviation from QAR records，while median errors in longitude，latitude，and altitude are 0.000 11°，0.001 2°，and 19.24 m，respectively，all satisfying safety separation requirements. Critical parameters including position，time，and heading angle at selected waypoints showed zero deviation. Notably，strict control of 13 waypoints is achieved within a 65-minute flight segment.

To comprehensively and efficiently analyze and evaluate chemical process risks，a risk analysis and evaluation auxiliary system for chemical processes was designed and developed based on SDG and CBR. The system was integrated with foundational verification functions for layer of protection analysis (LOPA) and safety integrity level (SIL)，enabling semi-quantitative risk analysis and evaluation. Additionally，CBR technology was utilized to effectively preserve and reuse historical case experiences in chemical processes，thereby enhancing the efficacy of HAZOP and the auxiliary evaluation system. The system was applied to conduct a risk evaluation of the propylene oxidation to acrylic acid process. The results demonstrate that the system aligns closely with the independent risk evaluation outcomes of the enterprise，exhibiting strong practicality and reliability. The SDG-CBR coupling reasoning mechanism is verified to overcome the reliance of traditional HAZOP on manual experience，achieving a qualitative-to-semi-quantitative transformation in hazard identification while enhancing evaluation efficiency and accuracy.

In order to improve the evacuation efficiency of public buildings and reduce casualties caused by emergencies，in view of the optimization problems of emergency exits and evacuation plans，a mathematical model and algorithm for the single-source multi-sink evacuation problem were proposed. Firstly，all evacuation paths were identified using path algorithm and sorted based on evacuation time. Then，the recursive formula for calculating evacuation time was derived based on the User Equilibrium (UE) principle and k-shortest path. The capacity allocation issues of each evacuation path at intersections were analyzed. After that then the optimization strategy for emergency exits is obtained. Finally，taking a building as an example，the evacuation network was established and calculated to validate the effectiveness of the algorithm and optimization strategy. The results show that the evacuation efficiency of the building network increases gradually with the increase in the number of evacuees，and the rate of change stabilizes after reaching a certain threshold value，Evacuation efficiency is closely associated with the total traffic capacity of specific emergency exits. Through the optimization of specific emergency exits，the evacuation efficiency of the network is enhanced，evacuation time can be reduced，and the optimization effect becomes more pronounced as the number of evacuees increases.