In order to prevent fire accidents in subway stations and deeply understand the internal relations and hierarchical structure among the influencing factors of fire evacuation in subway stations，63 fire evacuation accidents in subway stations were analyzed based on the sixth edition of 24Model，and the interaction among various factors was fully considered，and 19 key factors affecting fire evacuation in subway stations were extracted to establish an index system of influencing factors of fire evacuation in subway stations. C-OWA was used to improve DEMATEL to determine the important influencing factors of fire evacuation in subway stations. On this basis，ISM was used to analyze the hierarchical structure and interaction path of various factors，and a multi-level hierarchical structure model of the influencing factors of subway station fire evacuation was constructed. Research results show that evacuation guidance，panic herd behavior and crowded people are the key influencing factors for the evacuation of people in subway station fires. The evacuation of fire personnel in the subway station is influenced by surface factors，intermediate factors and deep factors. Among them，evacuation education and training，facilities maintenance and inspection，evacuation plan and other factors are the root causes. Paying attention to the improvement of root causes was conducive to preventing and controlling accidents in essence.

In order to further ensure the safety of railway operation，this article conducted a survey and analysis on the fatigue factors of railway train dispatchers. Firstly，according to the characteristics of the work content of railway train dispatchers，a fatigue causation evaluation index system for railway train dispatchers was established based on extensive literature review，consultation with railway train dispatching experts and human factors engineering experts. The index system consists of four first-level indicators: work content，job rotation arrangement，personnel quality，and external factors，as well as 18 second-level indicators. Secondly，AWAHP method was used to calculate the weight of different experts' own evaluation were calculated and the final weights of the final indicators were obtained. Finally，the dispatch center of a certain railway bureau was taken as an example for a comprehensive evaluation of the impact weights of dispatcher fatigue-cause factors based on AWAHP. The results show that emergency response，operation time period，continuous operation time，railway busy degree，and work complexity are the significant factors affecting dispatcher fatigue. The key factors of railway train dispatcher fatigue were effectively identified，and analysis and countermeasures were proposed for the key factors.

In order to further improve the prediction accuracy of slope stability，a slope stability prediction model based on MISSA optimized SVM was proposed. Six representative indexes，including bulk density (γ)，cohesion (c)，internal friction angle (Ф)，slope angle (φ_f)，slope height (H) and pore pressure ratio (r_u) were selected as the prediction indexes of the model. In response to the problems of slow convergence speed，low accuracy，and susceptibility to local optima in the sparrow optimization algorithm (SSA)，strategies such as one-dimensional composite chaotic mapping，SCA，Levy flight mechanism，and dynamic adjustment of step size factor are introduced for optimization and improvement. A slope stability prediction model based on MISSA-SVM was constructed. The MISSA-SVM model was applied to 9 groups of slope engineering examples，such as the Daxi landslide，for verification. The results show that the accuracy，precision，recall，F₁ score，mean square error (MSE) and area under the curve (AUC) of the MISSA-SVM model reach 96.29%，92.3%，100%，0.96，0.016 and 0.967，respectively，which are better than the SSA-optimized SVM model and BP model，and the prediction results are completely consistent with the actual slope conditions，indicating that the MISSA-SVM model has strong generalization ability.

An early prediction and warning method of offshore drilling overflow based on data model collaboration was proposed to prevent blowout accidents during offshore drilling. Firstly，an overflow risk prediction model based on PSO-LSSVM was established to predict the trend of drilling monitoring parameters in the future，and analyze the correlation between overflow events and characterization parameters. Then，a single-parameter overflow probability estimation prediction model was proposed based on the Naive Bayesian method，and the probabilities of multiple drilling parameters were integrated through the optimized D-S method to realize a hierarchical early warning of overflow events. The results indicated that the overflow characterization parameters simulated by the PSO-LSSVM model had low prediction errors. The overflow event probability represented by a single drilling parameter showed discrepancies due to different sensitivities. The fused early warning model can address the issues of inconsistent early warning times of single parameters and eliminate the possibility of false alarms.

In order to assist tunnel fire risk management and decision-making，the definition and principles of social risk acceptance criteria for road tunnel fire were established. Based on a statistical analysis of tunnel fire fatalities in China from 2000 to 2020 and a survey of public perceptions of tunnel fire risk，social risk acceptance criteria of tunnel fires in China was developed by the F/N curve method，and the acceptance critical limits for accident frequency and consequences were defined. The results show that the public has a tolerable attitude towards the current situation of tunnel fire accident risk. The maximum acceptable frequency of tunnel fire fatalities in China is set at 10^-5/(km·a)，and the maximum tolerable frequency is set at 10^-4/(km·a).

To face the challenges of evacuation and rescue of fires in large hydraulic tunnel group and the unclear control mechanism of smoke barrier，numerical simulation was employed to analyze smoke spreading law and fire control mechanism of smoke barrier. The smoke spreading path and the visibility distribution were analyzed. The impact of installing smoke barrier walls on smoke spreading speed，visibility and temperature was studied，and the effectiveness of smoke control measures was evaluated. The results show that the smoke initially spreads along a single two-way direction，and is affected by ventilation airflow when reaching branch tunnel with smoke reversal，retrogression and increased intrusion range. Smoke fills the whole ventilation network within 500 s，the stratification interface between the smoke layer and the lower air layer is unstable，and the height of the smoke layer at the intersection decreases more significantly. The smoke barrier can promote smoke propagation along the diversion tunnel，and the effect decreases with the increase of propagation distance. It has cumulative blocking characteristics along the tunnel direction of the construction branch. The influence of the smoke barrier leads to the temperature decreasing as the diffusion length increases. A smoke storage pool is formed in front of the smoke barrier to enhance the degree of backflow of smoke. Behind it is the smoke deceleration and pressurization area，and the smoke layer rises.

To deeply explore the knowledge production status and development logic in the field of risk communication for public health emergencies in China，this study systematically sorts out 234 core literatures from CNKI using bibliometrics and content analysis. First，the knowledge production trend is summarized through quantitative statistics; second，the Citespace software is used to construct a visual knowledge map of the main subjects and carriers of knowledge production and core issues，analyzing the current situation of knowledge production; finally，the content logic and process logic of knowledge production are sorted out through content analysis of the literature. The results show that the current knowledge production of risk communication for public health emergencies in China is driven by both "push" and "pull" factors. There is an obvious trend of interdisciplinary crossover，the content system of knowledge production is taking shape; and the methods of knowledge production are characterized by diversity.

In order to effectively prevent the unsafe driving behavior of take-away riders and reduce the traffic accidents caused by take-away riders，based on the work mechanism of take-away platforms' "rush game"，combined with the theory of stress perception evaluation，a theoretical model characterising the effects of time pressure and competitive atmosphere perception on unsafe driving behavior of take-away riders was constructed from the perspective of safety attitude and exemplary norms. A reliable questionnaire was designed through literature review，and 360 valid samples were collected from take-away riders all over the country. The data reliability was analyzed by SPSS 27.0 and AMOS 24.0 data analysis software，and the theoretical model was verified by hierarchical regression analysis using Process 4.0. The results show that time pressure and competitive atmosphere perception have significant positive effects on unsafe driving behavior. Safety attitude can significantly negatively affect the unsafe driving behavior of take-away riders. Safety attitude can mediate the relationship between time pressure and unsafe driving behavior，and also mediate the relationship between competitive atmosphere perception and unsafe driving behavior. Exemplary norms can positively regulate the relationship between safety attitude and unsafe driving behavior.

In order to improve the pilot's operational comfort，a combined evaluation method based on FPN and FRA was proposed. First，a multi-factor coupled FPN network topology was established for the uncertainty information of the influencing factors of operational comfort. Then，the optimal combination weights were determined by the game theoretic assignment method. An algorithm combining the inter-layer correlation to determine the critical importance，analytic hierarchy process and fuzzy reasoning was proposed to establish a FRA based on subjective and objective weights. The confidence and state matrices were solved by iterative solution of the library. Finally，combined with the scenario data of ship pilotage in Shanghai harbor，the operational comfort of pilots was evaluated through the application of FRA based on FPN. The results show that the environment and pilotage equipment are the key factors affecting the comfort of their operation. The comfort level of pilotage operation in winter and summer corresponds to "less comfortable"，and "more comfortable" in May. The proposed method fully reflects the coupled nature of the factors influencing the comfort of the system.

To quantitatively evaluate the influence of snowfall on the resilience of airport infrastructure system，the airport infrastructure system was divided into four first-level indicators，including physical system，functional system，economic system and organizational system. The damage intensity coefficient of snowfall weather to the second-level indicators was put forward，and a quantitative analysis model of airport infrastructure system resilience was established. The variation of resilience level of airport infrastructure system was simulated by the Monte Carlo Method under four different levels of snowfall. The simulation result show that the functional damage degree and recovery time of the airport infrastructure system are positively correlated with snowfall intensity. The system recovery time is 13，14，16 and 23 respectively under the conditions of light snow，moderate snow，heavy snow and blizzard. The resilience of the airport infrastructure system under the condition of blizzard is 9.65%，17.01% and 20.27% lower than that under the conditions of heavy，medium and light snow，respectively. The recovery speed of the airport infrastructure system increases from 0.053%/h to 0.061%/h，and the recovery time reduces from 12 h to 9 h when the failure strength coefficient reduces by 20%.