An Analysis of Runway Accident Precursors Based on Latent Class Model

Abstract

In aviation safety, there has been an increasing emphasis on accident precursor analysis. The analysis of accident precursors can provide the advantage of identifying factors that may lead to an accident before it occurs. This study analyzes rejected take-off (RTO) incidents, which is one of the runway accident precursors, using 326 RTO incident cases from 2011 to 2017. The combined dataset is constructed for RTO incidents that occurred in South Korea, by integrating incident report data, airport operational data, and aircraft registration data. The risk factors to RTO incidents are examined, including surrogate measures of time pressure. The two-step modeling is adopted to identify contributory factors. First, latent class model is applied to group similar incidents, and then negative binomial regression per cluster is conducted. The cluster-wise analysis provides a richer set of results in terms of commonality and diversity in the limited number of incident cases. Airline category and aircraft type are identified as the common contributory factors. The cluster-specific factors are congested airport, surrogate measures of time pressure, and the first flight of the day. The findings of this study provide in-depth implications on safety management to make further improvements in aviation safety in South Korea.