From raw operational flight data to incident probabilities using subset simulation and a complex thrust model

With the introduction of rules for safety management systems of airlines the need for providing incident probabilities exists. Utilizing operational flight data from Quick Access Recorders (QAR) and other data sources, the Institute of Flight System Dynamics at the Technical University of Munich developed a tool chain to compute incident probabilities for physics driven incidents. This includes decoding of binary data in ARINC 717 format, trajectory reconstruction using Rauch-Tung-Striebel smoothing, time point and measurement computation, thrust models partly based on GasTurb software and the subset simulation technique. For the incident type Runway Overrun conditional probabilities for certain sets of aircraft system states and runway condition, called causal chains, are computed in a subset simulation. This is done for different aircraft types landing at different runways. Operational differences between runways and runway conditions can then be taken into account in a subset simulation per causal chain.