A Hierarchical Probabilistic Approach for Risk Assessments of

Bayesian Networks (BNs) are widely used for knowledge representation and probabilistic inference in stochastic domains. An attempt is made in the risk modeling context to develop a general Aviation System Risk Model (ASRM) [16]. Following this initial ASRM, an operational/maintenance ASRM prototype is developed that is based on real case studies [17]. Past and current research efforts focus on multiple case designs in the context of an aviation accident category. Hierarchical Bayesian Networks (HBNs) are an extension to the conventional BN methodology, which provide an expressive power by allowing a node in the network to represent an individual BN. HBNs express the knowledge in a structured way, leading to more realistic probabilistic models. HBNs, like classical BNs, represent a stochastic modeling language, where the random variables symbolize the domain attributes. The probabilistic dependencies among the variables are modeled using directed acyclic graphs. A new case that combines different cases to obtain a single safety enhancement value can be designed using HBNs. The use of HBNs presents a novel application in the aviation safety domain and offers significant promise for risk assessments of an aviation safety product portfolio.