Predicting the Incidence of Military Aviation Safety Mishaps
Abstract
U.S. Air Force instruction requires commanders to conduct annual Mishap Prevention Analysis and develop safety plans that reverse adverse trends and guide prevention actions. Current analysis methods often fail to identify emerging trends in a timely manner and lack the ability to forecast the rate and trend of future mishaps based on historical data, resulting in arbitrary mishap rate reduction goals for the coming year. To properly safeguard service members and government property, commanders and safety staffs require timely trend identification and incidence forecasting provided by sound data analytic processes to target and combat safety trends. Using Air Force Safety Automated System (AFSAS) mishap and total flying hour data, we construct Time-Series Seasonal Decomposition and ARIMA predictive models that provide monthly trend, seasonality, and natural occurrence rates of Class A and B aviation mishaps and forecast the by-month mishap incidence rate for the following year. Each model provides valuable safety planning measures of performance. The seasonal model identifies months of peak incidence for safety plan targeting. The fluctuations of the ARIMA model establish the Band of Normal that identifies whether trends are exceeding natural rates. We develop a tool designed to be used by safety personnel at echelon to perform this analysis and predictive modeling, enabling them to focus on safety plan development instead of expending precious manhours processing data.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2022
- Accession Number
- AD1213795
Entities
People
- Ryan C. Herrmann
Organizations
- Naval Postgraduate School