A Bayesian network-based framework for risk analysis and assessment of unmanned aerial vehicle operations

With the rapid development of the low-altitude economy, the operational safety of unmanned aerial vehicles (UAVs) has become a major concern limiting their large-scale application. To address the complex risks caused by the coupling of environmental, technical, and human factors, this study proposes a UAV operational risk assessment framework based on the Systems-Theoretic Accident Model and Processes (STAMP) and Bayesian Networks (BN). The STAMP method is first applied to analyze UAV accident reports, identify potential risk factors and unsafe control actions (UCAs), and reveal interactions among human, machine, and organizational elements. These factors are then mapped to BN nodes to represent probabilistic dependencies, enabling quantitative inference and propagation analysis through expert knowledge and accident data. Model validation highlights critical risks such as structural failure, electronic speed controller malfunction, and inadequate pilot training, while sensitivity analysis indicates that organizational deficiencies and human errors most strongly affect accident probability. The results verify the framework’s effectiveness in UAV risk identification, dynamic assessment, and early warning, offering quantitative support for safety management and decision-making.