Oxidized graphene can be more rigid or flexible than graphene depending on the functional group and bending direction. Above a critical bending curvature, transition from O⋯H hydrogen bonds to O–H covalent bonds can occur in the oxidized graphene.

In recent years, the Arctic region, with its abundant oil and gas resources, has become a new focus of global resource development. However, the complex natural environment, especially the effect of sea ice, poses a serious threat and challenge to the navigation safety. Accordingly, this paper focuses on the navigation risks of drilling ships in five sea areas of the Northeast Passage of the Arctic under the influence of environmental factors. A dynamic Bayesian network structure was established using the Interpretative Structural Model–Bayesian network method. Since some risk elements cannot be directly measured, the combined weight method is adopted to fill the sample data. The navigation risk situations of the five sea areas is analyzed by forward causal reasoning. Through reverse diagnostic reasoning, the main risk factors affecting navigation are obtained, and relevant suggestions are given. This has important implications for improving the ability of accident prevention and emergency handling in practical applications. The model was verified through instance verification based on scenario analysis and model verification based on sample data. The average accuracy rate of the obtained model is 83.4%. The results show that the model has certain validity and practicability in the analysis of navigation risks in Arctic shipping routes.

In the extreme polar marine environment, safety risks pose a significant threat to drilling vessels. By conducting a safety risk assessment, potential hazards can be predicted and identified, thereby significantly reducing the frequency of accidents and promoting the sustained stability of economic activities. This paper investigates a Bayesian-network-based risk assessment model for polar drilling operations. Grey relational analysis was employed to identify the main risk factors. The model is trained using 525 valid incident sample data and is combined with expert knowledge. The accuracy rate is above 88%. Additionally, corresponding decision-making recommendations are provided through sensitivity analysis. The three most sensitive elements to fire nodes are human error, other causes, and equipment damage, with sensitivity coefficients of 0.046, 0.042, and 0.022, respectively. In terms of deck/handrail collision nodes, the highly sensitive elements are related to lifting (totally more than 0.1). For the events that have already transpired, the probabilities of most related nodes are 0.73 and 0.74, both of which are above 0.5, thereby validating the accuracy of forward and backward reasoning. Risk assessments based on Bayesian networks can offer pertinent decision-making recommendations and preventive measures.