A Framework for Uncertainty Assessment in Event Tree Safety Models

Abstract

The Integrated Safety Assessment Model (ISAM), developed by the Federal Aviation Administration (FAA), models aviation accidents and incidents via event sequence diagrams (ESDs) and supporting fault trees. Given available data from incident and accident reports, certain parameters in the event tree can be quantified. Yet, because many events are rare and are quantified with a small number of observed events, there is inherent uncertainty in the point estimates for the parameters. Further, some parameters in the model are not quantified at all, which contributes to the quantification uncertainty. This paper discusses a framework and methodology for quantifying uncertainty in the event-tree probabilities. The method is designed to be flexible in case that new data become available for parameters in the tree. Any new data can easily be added to update the uncertainty distributions and parameters. Several event trees with different levels of data availability are analyzed to illustrate the method and to identify the parameters with the highest uncertainty. ESDs with a large numbers of unquantified events exhibit more events with a higher level of uncertainty. We also introduce examples to show what changes to expect when new data are added in an ESD.