Analysis of accident precursor data for Mid Air Collision occurrences using expert-build Bayesian Network model and Information Theory

Abstract

Causes leading to loss of Separation (LOS) in serious and major incidents are considered as potential precursors for Mid-Air Collision (MAC) accident. This paper attempts to model the likelihood of these precursors combining Bayesian Networks (BN), which are based on expert-built, and Information Theory (IT). BN provides the analysis of LOS contributing factors and the multi-dependent relationship of causal factors identified from real Air Traffic Management (ATM) incident reports, while IT contributes to the identification of LOS precursors providing the most information. The combination of these two techniques allows us using data on causes and precursors of LOS to define warning scenarios. These precursors could forecast a serious LOS with severity A and B, and consequently the likelihood of a MAC. The methodology is illustrated with a case study that encompasses the analysis of LOS severity A and B that have been notified within the Spanish airspace during a period of four years.