Opinion: AI Should Drive The Future Of Aviation Inspections

Safety in aviation is paramount, and inspections have long been the critical line of defense against potential failures. Traditionally, these inspections have relied on human expertise—skilled technicians meticulously examining aircraft components for signs of wear, damage or potential malfunction.

The current inspection landscape has been established over decades and works in keeping high levels of safety in aviation. The regulatory backing to the current maintenance regime means the processes are intact.

There are headwinds facing the aviation industry. According to Boeing’s Pilot and Technician Outlook for 2024-43, “Long-term demand for newly qualified aviation personnel remains strong … 716,000 new maintenance technicians … will be needed to fly and maintain the global commercial fleet over the next 20 years.”

Combining advanced analytics on data from aircrafts has been a benchmark in aviation for over two decades now. And whilst engine monitoring is well-established, other critical systems lack the same level of attention. Visual checks, walk-arounds and working at height are still the norm. These happen so often that they are not questioned.

With the challenges ahead, I believe there is room for change.

Providing artificial intelligence (AI) assistance to highly valuable technicians and engineers in aviation needs to start now, before ultimately letting AI replace some elements of human inspection.

Will AI replace human inspection?

At Aviation Week Network’s MRO Europe 2024 conference I hastily said yes, without adding context.

I believe that AI will replace some human inspection, without putting a timeline on it, yet.

I entirely believe that by the end of this decade, AI will augment some human decision making. And humans will be involved in validating AI decisions rather than making the decision themselves.

Elements of AI are currently already in use. They are used to find instances of “What more can be done?” For example, power-by-the-hour programs by various engine OEMs use machine learning to find where there might be emerging issues.

Flying an inspection drone around an aircraft to capture images or video streams for real-time analysis is already being experimented with. Having AI that can review this data in the future is entirely feasible.

When the AI flags abnormality, it can hand over the inspection view to a human. This means the human is looking at 10% of the most valuable data, rather than getting fatigued by looking at 100% of the data when most of the data shows no issues.

A Dutch company, Aiir Innovations, delivers borescope analysis that it says is correct for the first time 50% more often than traditional inspections. It also has a ‘human-in-the-loop’ system where the final report is analyzed by a trained human and decision making is human-led.

In aviation, the adage “If there isn’t a problem, don’t fix it,” is well understood. But that mindset brought us supply chain challenges, various engine and aircraft issues, and trust issues with parts.

With the general availability of AI for all, we can for the first time in a cost-efficient manner try to tackle the issues that will creep up on us.

Availability of technicians is going to shrink, even with strong training plans. Delegating some of the low-level work that computer systems are best suited for should not raise eyebrows, and that is how I see AI propagating in aviation.

We need to use AI to assist humans do the work we trust them with. Reviewing data, watching hours of video to spot a crack or moving people and metal around an expensive aircraft to look for a crack is not the best use of rare human talent.

Let the binary decision of good and bad be left to AI, which does not have the capacity of the brain, and let our brains add the context and nuance that we train ourselves for to judge what the AI is delivering.

If we can do this for long enough, we will undoubtedly build the data and processes for some inspection work to be entirely AI led. Right now, the goal is no to mimic the human brain in inspections. The goal should be to mimic the human eye and let us get the AI to ‘see’ what we see and assist us first.

Faizan Patankar is CEO and cofounder of Amygda in Derby, UK. He has been focused on aviation and artificial intelligence since 2011.