The business aviation community is eyeing the potential of artificial intelligence (AI) to improve safety, which has always been a core value of the industry.
“There’s still a lot of curiosity about AI and what it can do,” said Greg Jarrett, CEO at aviation business operations systems provider Stack.aero. “People are still trying to figure out what the value is. At this point, everything in this artificial intelligence world is experimental, and will likely be in an experimental phase for the next two or three years.”
Along with curiosity, many industry experts are bringing skepticism to the discussion. For now, the idea of applying AI to aviation safety is simply “crystal ball stuff,” said Mike Ott, a Royal Aeronautical Society fellow and Gulfstream and Learjet captain with the Phoenix Air Group.
“They have a hammer and they’re looking for nails,” said Ott. “AI is a tool in its developmental stage. We are looking for uses of this tool to help guide its development within our industry.”
The definition of AI and its applications have yet to find focus, said Jeff Mittelman, a 40-year aviation veteran and longtime member of the NBAA Safety Committee.
“I’m somewhat of a critic,” Mittelman said, as technology that’s been common in aviation for years is now being called AI. Some types of flight operations quality assurance analyses serve as an example of where AI has long been applied – depending upon the precise definition of AI, he said. Mittelman believes AI’s real benefits will be in conjunction with machine learning – a related emerging technology.
Current Applications
AI “definitely” has a place in business aviation safety, said Rob Mather, vice president, aerospace and defense industries at Canadian software provider IFS.
“It can be applied today,” he said. IFS boasts clients ranging from the venerable engine manufacturer Rolls-Royce to advanced air mobility eVTOL developer Joby Aviation.
There is, to date, no organizational infrastructure to apply AI to aviation. “AI as we know it today relies on LLMs,” Jarrett said – large language models. Only big companies have their own LLMs, he said, so if you use Microsoft’s OpenAI/ChatGPT, for example, you’re effectively putting your proprietary data into the public domain – which may not be the best business decision.
Rolls-Royce, Mather said, is using AI, supported by data from IFS, “to help airline customers automatically update predicted maintenance deadlines for every life-limited engine component – what they call true-lifing.”
“Even though the FAA is a customer of ours,” said Mather, “I don’t know how they’re actually utilizing AI inside their walls. I do know that they should be.”
‘Viable Possibilities’
“There are viable possibilities as it becomes a more mature technology,” said Ott.
Experts agree that AI promises to take data analysis to new levels. “We are in the process of assembling an enormous database from flight data monitoring programs,” Ott said, “hundreds of parameters each second.
“People tend to react when things go wrong,” said Ott. “AI is going to give us a chance to discern the differences between events that go right and events that go wrong.
“AI can give us a better grip on what the flight data is really saying,” he added. “It’s a much more reliable form of flight data analysis than what we currently have. There’s an enormous pool of data pertaining to things that have gone right that we haven’t really looked at very closely.”
A Virtual Co-Pilot?
Researchers at the Massachusetts Institute of Technology are also using AI to parse data – in their case, visual data gathered during flight. Air-Guardian, developed at MIT’s Computer Science and Artificial Intelligence Laboratory, results in a virtual co-pilot to assist an aircraft’s human pilots.
Air-Guardian “blends human intuition with machine precision, creating a more symbiotic relationship between pilot and aircraft,” according to an MIT summary document.
The technology uses eye tracking to take in what the human pilots see, and “saliency maps” to pinpoint where their attention is directed, the summary says. “Intricate algorithms” allow Air-Guardian to pinpoint early signs of potential trouble, “instead of only intervening during safety breaches.”
AI and Maintenance
On the ground, AI is beginning to be applied to aircraft maintenance, where it can help address such chronic bugaboos as distraction, fatigue, stress and complacency. Predictive maintenance is nothing new but can be greatly strengthened via the vast data analysis capabilities of AI.
The trick with AI chatbots, said Mather, is “to know how to talk to AI to get the kind of answers that you want.”
“The monitoring of safety is very much up to the human,” Jarrett said. “The function of AI is assisting that person.”
And while AI probably won’t help much with business aircraft flight scheduling, which is dependent on ever-changing customer needs, it will almost certainly help with maintenance scheduling – determining not only when but where a maintenance operation can best be carried out.
What May Be Ahead
Mittelman noted that the FAA-led ASIAS (Aviation Safety Information Analysis & Sharing) collaborative is working to apply AI and other advanced processing techniques in the future; however, some might consider the long-standing ASIAS predictive analyses of tens of millions of aviation safety data to already include AI concepts. In addition to nearly 50 commercial air carriers, ASIAS counts 158 general aviation and on-demand Part 135 air carriers, and 15 helicopter operators among its stakeholders.
Mather said current uses of AI in aviation such as predictive maintenance and anomaly detection differ from future possibilities.
“In the future, rigorous testing regimes or the ability to explain how a model arrived at a conclusion or outcome is going to be key to widespread adoption in aviation,” he said. “I am extremely excited about the potential of AI.”