Analysis of the Key Precursor Indicator Parameters of Runway Excursion Incidents Based on Density Clustering Algorithm

Runway excursion incidents at the landing phase, reflected in quick access recorder data(QAR), are highly correlated with parameters related to speed, direction, touchdown distance, time, and use of reduction gear. Based on routine operational measurement data(ROM), i.e., snapshot data, this study establishes a system for runway excursion precursor indicators. By using density clustering algorithm, this study proposes a data extraction method to identify abnormal values, and further analyses the runway excursion precursor indicators in abnormal data to determine the incident cause. The density clustering algorithm has its basis in the assumptions that cluster centers are surrounded by neighbors with lower local density and that they are at a relatively large distance from any points with a higher local density.