Advancing Towards Automatic Detection of Safety Events

December 2021
Topics: Aviation Safety, Aviation and Aeronautics
Stephen Szurgyi, The MITRE Corporation
Aviation paths across the map of the globe
Download PDF (930.52 KB)

The Federal Aviation Administration (FAA) manages the world’s largest, busiest, and most complex airspace in the world. More than 87,000 flights move through this airspace each day. These flights stay safe thanks to the continued collaborative efforts of FAA service providers and industry partners. The safety level of today’s flight operations reflects a continual improvement in performance, technologies, and procedures. The goal of all aviation stakeholders is to make this system even safer. 

Legacy approaches to safety provision have been compliance-based. This approach has been highly effective in getting us to the incredibly safe system we have today. However, the past decade has seen significant advancements in aviation concepts and capabilities. The aviation system is in a discovery and rapid innovation phase for incorporating new entrants, controls, and business models. This evolving aviation system, as well as the role and responsibility of the FAA in ensuring continued safety, requires risk-based safety management. As the National Airspace System (NAS) gains increasing complexity via integrated operations, capabilities are needed to gather safety intelligence on systemic risks and the efficacy of safety barriers. 
The generation of safety intelligence must occur continuously and proactively to provide maximum benefit. We must leverage innovative approaches that utilize artificial intelligence (AI) to identify safety issues and the underlying contributing factors (precursors) that lead to undesired aircraft states and system safety states. These AI-based technologies must support the discovery of safety issues through automatic identification of non-normal operational behavior. 

There are many challenges to achieving this vision: safety culture, policy implications, human trust of AI output, and technological hurdles. To overcome many of these challenges, as noted by FAA Administrator Stephen Dickson, the FAA must be better at transitioning insights from its research partners into operational use. Now is the time to take action and convert advanced analytics research into the generation and operationalizing of safety intelligence.


Publication Search