关键词: 贝叶斯网络(BN), 事件树, 配对进近, 碰撞风险, 敏感性分析

Abstract:

In order to assess the collision risk of paired approach to closely spaced parallel runways, a collision risk model based on fuzzy Bayesian and event tree analysis was developed. Initially, the paired approach procedure was delineated, and risk factors inherent in it were identified through the Failure Mode and Effect Analysis (FMEA) method. Subsequently, a Bayesian network model addressing hazards during the paired approach and potential control failures was constructed, leveraging the identified risk factors. The probability of the root node was determined through a combination of expert survey weighting and statistical analysis of historical data. Some root node probobilities and the conditional probability of the intermediate nodes were fuzzified utilizing seven-level linguistic variables, followed by de-fuzzification using the incentre of area. Additionally, priori probabilities and sample data were input into BN software for expectation-maximization(EM)parameter learning, facilitating the determination of hazard proximity and control failure probabilities. Lastly, considering the time series relationship between hazardous approach, control failure, and collision events, the collision risk associated with paired approaches was assessed employing event tree analysis, and a sensitivity analysis was conducted on the BN. The result shows that hazardous approach sensitivity is highest to pilot operating level, while control failure sensitivity is highest to poor maintenance. If the probability of poor pilot performance surpasses 12%, and the likelihood of inadequate maintenance surpasses 0.17%, the paired approach operation fails to meet the safety target level.

Key words: Bayesian network (BN), event tree, paired approach, collision risk, sensitivity analysis