Bayesian network modeling of accident investigation reports for aviation safety assessment

Safety assurance is of paramount importance in the air transportation system. In this paper, we analyze historical passenger airline accidents that happened from 1982 to 2006 as reported National Transportation Board (NTSB) aviation accident database. A four-step procedure formulated construct a Bayesian network capture causal relationships embedded sequences these accidents. First all, with respect each accident, graphical representation developed facilitate visualization escalation initiating events into Next, develop all by aggregating accident-wise representations together, where and dependent among wide variety contributory factors outcomes terms aircraft damage personnel injury are captured. network, prior probabilities estimated based on occurrence times departure data Bureau Statistics (BTS). To estimate conditional monotonically increasing function, whose parameters calibrated using probability information single available data. Finally, computer program automate generation compliance XML format used commercial GeNIe modeler. The constructed then fed modeler for analysis. mapping NTSB facilitates both forward propagation backward inference probabilistic analysis, thereby supporting investigations risk Several cases demonstrate approach.