Application of a temporal Bayesian network for safety risk analysis

An imperative step of the aviation safety risk management process is to establish a method that evaluates the likelihood of undesired events. In complex domains such as aviation safety, where low probability, high consequence events are prevalent and where data are scarce, the range of effective risk analysis tools is limited. There is a need to develop a comprehensive analytical method that captures the explicit and implicit risks inherent in complex domains. A Value-based Time-phased Bayesian Network (VTBN) is an extension to conventional Bayesian Networks (BNs). They incorporate features from Dynamic Bayesian Networks with temporal factors and are enhanced with the addition of Multi-Attribute Value Theory (MAVT). The VTBN integrates the quantitative constructs of BNs and MAVT with the qualitative formalisms of hazard taxonomies. This enhanced method allows the modeler to capture the explicit risk apparent from the BN and the implicit risk not apparent in the BN that is prevalent in large complex systems. Initial modeling results suggest VTBNs offer significant promise for advanced risk assessment, particularly for Unmanned Aircraft Systems (UAS). In this paper, the analytic constructs of a VTBN are demonstrated with an application of safety risk modeling for a hypothetical UAS scenario. This research is supported by Federal Aviation Administration grant number 08-G-002. The contents of this paper reflect the views of the authors who are solely responsible for the accuracy of the facts, analyses, conclusions, and recommendations represented herein.