The event tree analysis(ETA) was applied to the risk propagation process of the risk event chains involved in ship collisions for the purpose of developing directed weighted complex，which was then analyzed according to the topological structure. The traditional K-shell decomposition algorithm was adaptively improved to propose the calculation approach for the absolute criticality of the nodes in the CN. The SIR model was utilized to simulate the risk propagation process within the CN，based on which the dynamic sensitivity of the nodes within CN was analyzed under different cases of infection rates and recovery rates. It is pointed out that approximately 25% of the risk events are found as relatively non-active，and the top 10 critical risk events have no relation with the direct causes of the collision accidents. The sensitivity and criticality of risk events are not consistent at all times，and some risk events would not lead to serious consequences or accidents even though countermeasures are not taken. The existence of direct causes for the collision accidents should be acknowledged for the control and management of risk involved in the ship collisions，and much attention should be paid to the risk propagation process. As a result，differential strategies should be taken during the risk management and control to prevent the ship collisions.

In order to improve slope stability under the combined action of engineering disturbance and natural factors，and clarify the process and characteristics of slope soil nailing under different loading conditions，the optimal support mode and layout technology suitable for the actual working conditions were selected to ensure the intrinsic safety of the slope structure system. Firstly，the physical model test process and results of soil nailing support under three types of static，vibration，and centrifugal forces were systematically elaborated. Secondly，the characteristics and applicability of typical slope support technologies and new slope support methods were compared as well as analyzed. Finally，through physical model tests under complex environmental effects，the deformation and failure mechanisms of three special slopes，namely submarine slope，typhoon rainstorm slope and high-speed and long-distance landslide were explored. The results show that soil nailing has good applicability in all kinds of slope support，but the reliability and safety factor of slope support under complex and special environments still need to be improved. Therefore，in practice，it is necessary to integrate new materials and processes to conduct research on composite support structures. At the same time，it is necessary to strengthen the selection of similar materials，innovative observation methods，and special environmental simulation research in physical model experiments，aiming to improve the high degree of restoration of monitoring data and achieve the full process safety guarantee of slope systems.

In order to clarify the relationship between the exterior insulation facade structure and fire spread in high-rise buildings，Pyrosim fire simulation software was used to study the impact of different exterior insulation facade structures on fire spread in high-rise buildings. The results show that during the process of fire spread on the exterior facade of high-rise buildings，the insulation systems of different exterior facade structures reduce the air entrainment capacity and heat release rate as the degree of structural space limitation increases. However，their chimney effect is significantly enhanced，and the smoke flow rate is faster，leading to an accelerated vertical fire spread speed. As the thickness of the air layer increases，the temperature and smoke flow rate of the aluminum curtain wall structure insulation system first increase and then show significant fluctuations. The aging of the performance of the external wall insulation system will increase the risk of fire on the exterior facade of high-rise buildings.

In order to study the smoke transport characteristics of rail train with lateral multi-opening and optimize the emergency response strategy for fire rescue，a rail train model was built using PyroSim software based on the engineering entity structure and numerical simulation was conducted. Focusing on the lateral opening as a connecting channel between the inside and outside of the train carriage，the influence of the lateral opening mode on the smoke transport characteristics inside the train was discussed and analyzed. The results show that the lateral opening mode has a significant effect on the smoke transport inside the train，including the smoke spreading speed and the smoke distribution in the stable stage. Compared with the disordered smoke and full coverage in the height direction in the section without opening the lateral opening，the smoke in the section with opening the lateral opening can maintain stable stratification，which is conducive to the safe evacuation and rescue of personnel. The smoke temperature distribution under the train ceiling is more consistent with the double exponential attenuation law，and is affected by the coupling of the opening mode and the fire power. The temperature attenuation rate of the opening section is more significantly affected by the fire power in the asymmetric opening. In addition，although the opening is conducive to the formation of a safe environment inside the train，it is also necessary to consider the impact of the fire smoke spilling into the outer space of the train through the opening to make a comprehensive judgment of emergency response.

To solve the issues of wind turbine blades in terms of classification difficulty and blurry segmentation of small defects in surface defect detection，an improved U-Net semantic segmentation network was proposed based on dilated convolution and convolutional attention modules. Based on the encoding-decoding structure of the network model，a transferable VGG16 feature extraction layer was used to replace the encoding part of the U-Net network. Then，a convolutional attention module was added to the skip module between encoding and decoding. The global weight was enhanced by selecting small defect information. Dilated convolution was used in the decoding section to enhance the network's feature extraction ability，and the pre-trained VGG16 model was used to realize transfer learning. The hybrid loss function of Focal and Dice was validated against the models of DeeplabV3+，Pyramid Scene Parsing Network(PSPnet)，High-Resolution Network(HRNet)，and U-Net. The results showed that the improved U-Net network had higher prediction accuracy in blade defect classification and segmentation tasks，mean intersection over union，mean pixel accuray，and recall values were 83.60%，92.84%，and 88.50%，respectively. The mean intersection over union simulated by the improved U-Net model was 13.98% and 9.38% higher than that by the DeeplabV3+ and standard U-Net model，respectively. Therefore，the proposed model can improve the sensitivity of blade defect detection，effectively reduce false positives of detection results，and provide guidance to wind turbine blade defect detection.

In order to ensure the navigation safety in the process of ship driving and operation and accurately evaluate the practical operation skills of the crew，aiming at the problem of the lack of a complete evaluation index system for the evaluation of ship mooring operation，combining the safety evaluation of mooring operation conducted by the expert evaluation method and fuzzy comprehensive evaluation method，the membership function of each evaluation index was constructed. The analytic hierarchy process(AHP) was used to determine the suggested weight value and the standard value of each evaluation index to ensure the accuracy and reliability of the evaluation results. The three-dimensional virtual ship technology was used to develop a set of mooring maneuvering automatic evaluation systems according to the influence degree of each factor on mooring maneuvering，so as to realize the comprehensive evaluation of the ship's mooring operation. The study shows that the evaluation model has better accuracy，strong systematization，and easy to operate. The results of the system's assessment，when compared with the results of the expert assessment，showed a high degree of consistency，demonstrating the validity and reliability of the model in assessing the mooring skills of crew members.

To reduce the influences of background interference factors in natural environments such as clouds，mist，dust，lights，sunrise，and sunset on the smoke and flame target detection accuracy，a smoke and fire detection algorithm based on an improved YOLO-V5 algorithm was proposed. Smoke，flame target images，and interference image data sets were obtained from the on-site collection and web crawling approaches to solve sample imbalance and improve model generalization ability. A bidirectional feature pyramid network (BiFPN) was used to replace the original feature pyramid network (FPN) + path aggregation network (PAN) structure，and then multi-scale feature fusion on the target was performed to strengthen the model feature fusion ability. At the same time，distance intersection-over-union(DIoU) non-maximum suppression(NMS) is used to replace the original non-maximum suppression (NMS) to speed up the convergence of the detection box loss function and enhance the model reasoning ability. The results showed that the improved algorithm's accuracy，recall rate，mean average precision(mAP) and FPR were 79.2%，68.6%，74.2%，and 12.8%，respectively. Compared with the original YOLO-V5 algorithm，the proposed algorithm improved accuracy rate，recall rate，and mAP by 1.9%，0.9%，and 2.7%，respectively. Furthermore，the FPR was decreased by 3.7%.

To effectively control the smoke spread in large flat-space ships，the oil pool mass loss rate，chamber temperature distribution，thermal insulation efficiency，and smoke control effect were investigated during large-scale fires. The experimental study was performed under three different smoke screen heights (0.35，0.55，and 0.70 m) and mechanical ventilation conditions in a chamber with dimensions of 30 m×24 m×2.3 m. The results indicated that the increase in smoke screen height caused a decrease in the peak value of the oil pool mass loss. The smoke screen height had a more significant impact on the smoke temperature in the chamber's upper layer than that of the lower layer. The peak temperature above 1.4 m was significantly decreased，whereas no significant change in the peak temperature below 1.4 m was observed. The average ceiling temperature and thermal insulation efficiency decreased with the increase in the smoke screen height. The thermal insulation efficiency increased from 28.2% to 50.8% under the combined smoke control mode，and it increased from 29.4% to 54.7% under the independent smoke control mode.

In order to solve the problems of difficulty in quantifying the indicators and difficulty in taking into account randomness and fuzziness in the evaluation process of the risk of the coal mine intelligentization project，the cloud model theory was adopted to carry out a quantitative and comprehensive evaluation of the system. First of all，based on the coal mine informationization system construction project，a multi-dimensional analysis was carried out to establish a multi-indicator and multi-dimensional evaluation system of the project risk. Then，the combination of hierarchical analysis (AHP) method and criteria importance though intercrieria correlation (CRITIC) method was used to assign weights，determine the weight matrix of the indicators，and the cloud model was used to realize the conversion between the quantitative and qualitative indicators，to complete the evaluation of the risk of the coal mine intelligence project，and to put forward the targeted policy according to the evaluation results. The cloud model was used to realize the quantitative and qualitative conversion of indicators，complete the risk evaluation of the coal mine intelligentization project，and put forward targeted policies based on the evaluation results to minimize the existing risks of the project. Finally，taking a coal mine of National Energy Group as an example，the risk evaluation of the construction and implementation process of coal mine intelligentization project was carried out. The results show that the cloud model can realize the quantitative evaluation of project risks，and the results of risk evaluation coincide with the actual situation on the site; the results of risk evaluation can help to solve the hidden risks on the site and improve the ability of risk control.

In order to explore the coupled evolution mechanism of air traffic operation safety risk，clarify the mechanism of coupling and mutation formation in air traffic operation systems based on a combination of the N-K model and FRAM. Firstly，textual data on unsafe incidents was collected. The risk factors involved were categorized，and their historical frequency of occurrence and the coupling relationship between risk factors were obtained. Secondly，the N-K model was used to solve the coupling degree values between air traffic operational risk factors. Finally，based on the output time and accuracy，the variability of the FRAM functional module was quantitatively evaluated，analyzing the coupling mechanism of air traffic operational safety risks，and safety risk analysis was conducted using regional area navigation(RNAV)approach unsafe events and deviation route unsafe events as examples. The results indicate that the evaluation method based on improved FRAM can quantitatively calculate the variability between functional modules in a reasonable and effective manner，weaken the dependence of traditional analysis methods on subjective consciousness，and make the analysis results more objective and scientific.