PNNL

Abstract

Algorithms have been developed to find typical patterns and atypical events within complex data systems. A software package called “The Morning Report” was developed in which these algorithms were applied to digital flight data for commercial airlines. These systems contain many sets of data with hundreds of variables being measured over time generally resulting in many gigabytes of data to be analyzed. Using statistical and mathematically based algorithms this software identifies atypical flights, along with identifying which flight parameters and which flight phases are atypical. These algorithms also cluster the flights into a finite number of distinct patterns. This allows the flight analyst the opportunity to focus on atypical flights, as well as the typical flight patterns discovered, removing the need to individually explore each flight separately. This software is titled “The Morning Report” because it was designed to run each night, producing a report in the morning. This report only identifies the characteristics of the newly added flights, but it uses past flight data to help establish the baseline. The report consists of interactive analysis tools that allow for plotting of significant flight parameters for each atypical flight as compared to the typical flights, as well as plots that contrast a flight pattern of interest to any other flight pattern, or all patterns combined.