Walkaround

Technology · On the aircraft

AI assisted inspection

Borescope image analysis, blade inspection, damage flagging

in service and spreading quickly 2 min read · updated 15 October 2026 · checked 15 October 2026

Two independent sources on every claim.

Why it is here

It targets the variability in how two competent inspectors read the same image, which is harder to plan around than a slow inspection.

At a glance

What it is

A borescope is a camera on a flexible probe, pushed into an engine through an access port so the internals can be inspected without disassembly. Getting the image has never been the hard part. Deciding what a mark on a blade actually is, and whether it is within limits, is the hard part.

AI assisted inspection does not make that judgement. It reviews the captured imagery and tells the technician which frames deserve attention, so the expert time goes to the frames that matter rather than to scrolling past the ones that do not.

In plain terms

The tool is a filter, not an inspector. It narrows thousands of frames down to the handful worth studying. A person still decides what the finding is and whether the part stays.

GE's tool has been running on the GEnx for roughly three years and is deployed across more than a dozen of the company's own shops, which is a longer service record than most tools in this category.

Why it matters on the line

On the line

The reported headline is time, but the number that should interest a planner is variability. Borescope interpretation is among the most experience dependent calls in the shop, and two competent inspectors can read the same frame differently. A spread in findings is harder to plan around than a long inspection, because it shows up as unpredictable removals rather than predictable hours. If a tool genuinely narrows that spread, it improves forecasting as much as throughput.

In plain terms

An unplanned engine removal costs far more than the same work planned, because the aircraft is already somewhere inconvenient, the part is not staged and the shop slot is not booked. Most of the value in better inspection is moving work from the first category to the second.

Judge an implementation on what it writes into your maintenance system rather than on the model behind it. A finding that arrives early enough to become a planned removal is worth a great deal. The same finding surfaced in a portal nobody opens is worth nothing.

The workforce argument is real and should be stated carefully. A great deal of interpretive skill is retiring from this industry, and a tool that encodes some consistency is a reasonable response to that. It is not a substitute for the judgement, and anyone selling it as one is overselling.

Where to read more

The story is GE and Pratt put AI in front of the borescope. The adjacent topic is health monitoring and predictive maintenance, which works on in flight data rather than inspection imagery. The engines named are the GEnx, the CFM LEAP, the GE9X and the V2500.

Sources