PNNL

Abstract

This paper presents a data mining process that is applicable to any number of domains with large amounts of data, and an interest in identify typical patterns, atypical events and trends over time. It is implemented in an aviation safety domain. Commercial airline safety has improved to the point that flying is the safest form of transportation available to the public. The strides in flying safety are the result of concerted efforts by the airlines, the government, and the aircraft industry-efforts largely based analysis of accident data. Any significant additional improvements in aviation safety will require a new approach, a new methodology, for analyzing the collective flying experience of the airline industry. The impetus for developing new tools to improve flight safety is the projected doubling in the number flights in the next few years and the concurrent two- to four-fold increase in the number of accidents. Researchers at Battelle working under a contract with NASA's Ames Research Center and directed by Irv Statler, are developing a statistical analysis tool to examine non-accident flight data to generate insight and understanding about the typical and atypical conditions that affect aircraft flights.