Categorization of Coast Guard Aviation Mishaps Following Aircraft Model Transitions: A Natural Language Processing Approach

Abstract

Aircraft model transitions at United States Coast Guard (USCG) air stations present a unique, and often hidden, set of hazards that may only be identified through reported mishaps. Although all USCG aviation mishaps are tracked, there is limited documentation and analysis of the specific factors that occur following an aircraft model transition. This lack of documentation and analysis contributes to the potential for future mishaps when similar aircraft model transitions take place. This study categorized reported aviation mishaps following aircraft model transitions through a quantitative analysis of mishap class, Operations Mode, type, and frequency. Additionally, Natural Language Processing (NLP) techniques were applied to identify latent mishap topics. Aviation mishap reports (n=422) for the three years following the first flight of a new aircraft model were extracted from the USCG’s Electronic Aviation Incident and Accident Tracking System (e-AVIATRS) for 11 USCG air stations which conducted transitions between 2004 and 2021. Results indicated that the greatest number of reported mishaps, as well as the most severe mishaps, occurred during the second year following the aircraft model transition. Mishaps relating to airframe or engine damage, as well as personal injury, were universally reported by all aircraft models examined and verified by potential NLP topic modeling. This study revealed the need for additional analysis of future air station transitions to identify latent trends for increased risk mitigation. A similar methodology could be applied to aviation mishap reports following airframe transitions or other significant operational changes at the state and federal level, as well as in commercial aviation.