The goals of carbon peaking and carbon neutrality support the development of HFCV industry. Safe parking of HFCV in outdoor three-dimensional parking lot is the key to large-scale promotion. Therefore, a scaled-down physical experimental platform was built to investigate the effects of outdoor multi-story parking lot closure conditions (closed, semi-open, open), leakage location, mass flow rate nd aperture on hydrogen diffusion characteristics. The experimental results show that as the opening degree of the outdoor multi-story parking lot increases, and the related indexes of helium at the leakage source rise. For example, when the leakage occurs at the center with a large aperture at the flow rate is 0.18 kg/s, the maximum volume fractions of helium on the first, second, and third floors are 21.3%, 27% and 45.3%, respectively. The corresponding hydrogen volume fractions far exceed the lower explosive limit, posing an extremely high risk of combustion and explosion. With the increase of leakage mass flow, both the volume fraction and pressure of helium increase, and the safety risk increases. There is an obvious positive correlation between leakage mass flow rate of helium and volume fraction and pressure. Under the same degree of openness, helium tends to accumulate in corners, and its diffusion along walls is more efficient than that in open areas. The diffusion process is guided by the spatial boundary, showing a non-uniform diffusion trend. In addition, the larger the pore size, the higher the helium leakage rate.

In order to investigate the influence of air pressure and ventilation speed on the flashover critical conditions of cable fires in long-narrow confined spaces under plateau environments, a functional equation between the flashover critical ceiling temperature and air pressure and ventilation speed was derived based on dimensional analysis. A cable fire model in a long-narrow confined space was established using Fire Dynamics Simulator (FDS) to study the critical flashover temperature during fire development under different air pressures (101, 90, 77, and 65 kPa) and ventilation speeds (0, 0.5, 1.0, and 1.5 m/s). The results show that the flashover critical ceiling temperature of cable fires in long-narrow confined spaces is relatively high, ranging from 720.7 to 877.3 ℃. The flashover critical ceiling temperature follows a power-law relationship with air pressure and ventilation speed, while showing a decreasing trend with increasing pressure and ventilation speed. The established model provides reasonable predictions for the flashover critical ceiling temperature of cable fires in long-narrow confined spaces under both natural ventilation and longitudinal ventilation scenarios. The model prediction results agree well with the numerical simulation results.

To deeply explore the underlying patterns of road traffic accidents involving Autonomous Vehicles (AV), relying solely on the statistical analysis of individual accident description factors was insufficient. It was necessary to uncover further the comprehensive latent categories reflected by the interactions of multiple factors. Given that AV accident data contained structured information and narrative text, an innovative approach was proposed for type identification combining K-means clustering analysis and LCA. Specifically, the K-means method was used to extract key information from the narrative text, which was then fed into the LCA model to overcome the limitation of LCA being able to utilize only the structured information in existing accident reports. The effectiveness of this combined approach was verified using 437 AV traffic accidents in California, USA. The results show that AV accidents mainly manifest in four comprehensive types. The combined approach of K-means and LCA enables efficient clustering analysis of structured information that includes narrative text.

In order to further enhance the early abnormal warning capability of natural gas compressor units, a novel method was proposed based on the VMD (Variational Mode Decomposition) algorithm, Informer algorithm, 3σ criterion, and Correlation-Weight optimization. A predictive model was constructed using the Informer architecture, in which the monitoring data were decomposed by VMD algorithm into multi-scale features of different frequencies to serve as model inputs. During the training process, the weight coefficients between each decomposed component and the original signal were calculated to optimize and adjust the internal model parameters. Furthermore, the prediction reconstruction results were combined with the 3σ statistical criterion to further improve the warning performance. Two segments of normal and abnormal pressure differential monitoring data from field compressor units were collected for experimental validation. The results show that, compared with other prediction methods, the proposed warning method achieves the lowest prediction errors. For the prediction of normal box pressure differentials, the errors are reduced by 66.67%-71.43% (Mean Squared Error, MSE), 36.67%-45.45% (Mean Absolute Error, MAE), 40.17%-45.42% (Root Mean Squared Error, RMSE), and 36.57%-45.72% (Mean Absolute Percentage Error, MAPE). For the abnormal inlet pressure differentials, the errors are reduced by 64.43%-71.12% (MSE), 44.02%-52.27% (MAE), 40.36%-45.53% (RMSE), and 37.24%-47.79% (MAPE). The proposed method exhibits superior prediction accuracy in both detailed and trend features. In the test set, it provides anomaly warning 60 minutes in advance, thereby improving the reliability of safe and stable operation of the compressor units.

In order to solve the problem of low guidance efficiency of emergency evacuation signs, a layout optimization method of emergency evacuation signs considering pedestrian perceptual behavior was proposed. Firstly, the recognizable range and probability distribution of the signs were investigated to establish a model of pedestrians' perceptual behavior towards the signs; secondly, a layout optimization model was constructed considering the number, location and orientation of the signs by combining it with the distribution of the guidance demand in the multi-exit space; finally, a case study of a cruise ship was used, where a genetic algorithm is applied to solve the problem. The evacuation efficiency, both before and after the optimization of the signage layout, was evaluated through microscopic crowd evacuation simulation, respectively. The results show that the average evacuation time under the optimized layout is 10.3% shorter than that under the unoptimized layout when the pedestrians are randomly distributed at the initial position; the average evacuation time is 4.2% shorter under the scenario simulating the actual evacuation process, and the number of pedestrians guided to different destinations is more balanced.

To verify the impact of drivers' evasive actions on injuries of two-wheeler riders and to identify the conditions under which evasive maneuvers fail or even exacerbate the injuries, based on data from China In-Depth Accident Study (CIDAS), NSGA-II was first used to simulate and optimize the reconstruction of accident scenes. Vehicle speed and two-wheeler speed data at the moment of collision were extracted. A research dataset consisting of independent variables, dependent variables, and 11 covariates was constructed. Simulation-derived variables and investigation-based data fields were integrated to serve as the foundational data for modeling. Secondly, two causal inference methods were adopted, namely the propensity score (PS) weighting - regression analysis combination method considering positive hypotheses and covariate adjustment (inverse probability weighting (IPW) and overlap probability weighting (OW)), to infer the causality between evasive actions and injury severity, and to compare the inter-group balance after processing by IPW, OW and unweighted regression methods. Finally, the causal effect of drivers' evasive actions on the injury severity of two-wheeler riders under different conditions was quantitatively analyzed. The results show that, in general, the evasive actions currently adopted by drivers cannot effectively reduce the injury severity of riders. When the vehicle types are commercial vehicles and the motor vehicle traveling at medium to high speeds, taking evasive driving actions tends to aggravates the injuries of riders. Among these actions, steering maneuvers is more likely to increase the severity of rider injuries.

In order to enhance safety in production, a human injury mechanism of operation was studied based on the principles of human BES. A safety model for human factors in operations was integrated from biological energy and the functional capabilities of biological tissues. The energies interacting with the human body were categorized into four types: the ingested energy, the activity energy, the substance energy in production and the energy from surrounding. The way they influenced the operation safety was analyzed. Based on injury consequence state, the characterization function of human injury degree was constructed, and mechanism of human injury under the action of substance energy was revealed. It shows that the abnormality of substances and energies in production operation produces overaction in the human body and leads to the blockage or imbalance of biological energy activities, which results in biological tissue damage and loss of function. This is the root cause of human injury. The substances and energies which interact with the human body come from external surroundings and internal (e.g. the body's biological energy). All the root causes of occupational health injury and production safety injury can be traced to energy interactions. That helps us to form a unified mechanism of human injury theoretically.

In order to provide a comprehensive overview of the current state of research and knowledge evolution in the field of occupational hazard factors and health risks associated with building decoration, a systematic review of the relevant literature was conducted using the China National Knowledge Infrastructure (CNKI) database and the Web of Science (WoS) core database. Visualisation tools were employed to make knowledge maps of the spatial and temporal distribution, research hotspots and research frontiers of the identified literature. The results show the number of published articles in both Chinese and English increased significantly since 2003. The relevant research is influenced by economic conditions, the real estate market, and the level of emphasis placed on health by both national and international communities. The extant literature can be broadly classified into three categories: workplace environment, occupational hazard factors and health risks. Studies have been conducted throughout the entire period pertaining to indoor air pollution, indoor air quality, detrimental factors and formaldehyde. The research field continues to yield new insights and deepen its understanding. However, there remain areas that require further investigation with regard to the occupational health of workers. In the future, the studies could focus on the occupational health problems caused by different types of decoration, with a particular emphasis on the pollution characteristics of new pollutants and the occupational health risks and protection of workers.

In order to enhance the precision of the criteria for identifying major accident hazards, an evaluation method for major accident hazards based on ATA was proposed. This method systematically applied evaluation tools such as the risk matrix and accident tree analysis to determine high-risk and above accidents through accident risk classification and used as the top event for ATA. The accident tree was simplified using Boolean algebra rules to obtain the ranking of the structural importance of each basic event that could induce the accident. According to the "80/20 Rule", the top 20% of basic events with the highest structural importance were identified to determine the core events that could lead to the accident, which were then regarded as major accident hazards for precise control. A case study on high-rise building fires was conducted for validation. The results demonstrate that the failure of the automatic fire alarm system has the highest structural importance in the fire accident tree of high-rise buildings through ATA. It should be judged as a major fire hazard clause of high-rise buildings. Therefore, the evaluation method for major accident hazards based on accident tree analysis can effectively determine the criteria for identifying major accident hazards.

To solve the problem that construction safety requirement information hidden in project documents is hard to be discovered without relevance and semantic ambiguity, a two-stage integration framework combining NLP techniques was developed for project document analysis and classification and extraction of requirement information. First, the safety targets of the project to be evaluated were obtained by combining the multivariate techniques of NLP, and an association model was established based on the topic model to recommend the appropriate requirement types. Then, the semantic features of the three types of elements were considered, and keyword analysis, sentiment analysis, and dependency analysis were adopted to extract the three types of elements, respectively. Finally, two types of construction projects (civil and industrial) were used as case to test the type recommendation and extraction of construction safety requirements. The results show that the two-stage integration framework recommends four appropriate requirement types for civil and industrial buildings respectively, and the combination of lexical properties and lexical sentiment can effectively extract the requirement keywords and behavior opinion words, and the extraction accuracy of the main elements can reach 88.6% after supplementing the description of building types. The test results confirm that responding to safety target features can recommend suitable types from the complicated requirement information, and the classification and extraction of requirement information combined with NLP avoids subjective preferences and improves the accuracy of information extraction.