EPILOTS: A SYSTEM TO PREDICT HARD LANDING DURING THE APPROACH PHASE OF COMMERCIAL FLIGHTS

Abstract

A go-around might have 
saved more than half of all commercial 
aircraft operating accidents. The total 
accident rate in the aviation business may 
be lowered by making the choice to do a 
go-around maneuver in a timely manner. In 
this study, we report on the development of 
a deployable machine learning system for 
the cockpit that facilitates go-around 
decision-making by the flight crew in the 
case of a hard landing. This paper provides 
a hybrid technique for hard landing 
prediction that feeds a neural network with 
features modeling the temporal 
interdependence of aircraft characteristics. 
The findings demonstrate that our 
technique has an average sensitivity of 85% 
and an average specificity of 74% at the go
around point, based on a large dataset of 
58177 commercial flights. Thus, our 
method—a cockpit-deployable 
recommendation system—performs better 
than previous methods.