Training AI to recognize paint-peeling

Training AI to recognize paint-peeling

Written by Alessandro Scoppio

Discover how Mainblades and KLM are changing aviation maintenance at Schiphol, Amsterdam, using AI-powered inspection drones to tackle paint peeling on their fleet. Leveraging Mainblades' automated inspection platform, visual inspections reach unprecedented speed and precision, setting a new standard in aircraft upkeep.    


The challenge of paint peeling

The composite material of modern aircraft such as Airbus A350s or Boeing 787 suffers from the issue of paint peeling, especially on the wings. This is not only an esthetic issue concerning passengers that approach the aircraft when onboarding, but also evidence of substantial degradation of the composite fiberglass ply covering the aircraft body. Such damage is a result of exposure to UV radiation and poses challenges to the aircraft's structural integrity. The short-term fix for such an issue as suggested by aircraft manufacturers is... high-speed tape.  

While the tape used is aviation grade speed tape, made specifically to withstand cruise speed of an airliner, this does not change the customers’ perception and concerns: seeing a wing patched up with square meters of what looks like regular duct tape will not alleviate their flight anxiety. Additionally, there is only so much speed tape that can be applied without affecting flight performance, and once a certain amount is reached per component, that component needs to be repainted.  

The way airlines have coped with this issue so far is to regularly perform visual inspections of the aircraft wings to ensure that the patches of peeled off paint are covered with speed tape, and to monitor the amount of tape applied. Pictures are usually taken manually using smartphones to assist in this task. Nevertheless, this kind of visual inspection can be quite costly for MROs, which must not only schedule and perform such checks – resulting in downtime for the inspected aircraft – but also provide specialized personnel to visually estimate the amount of paint peeling and tape applied. No matter how well trained and experienced the engineer is, this poses significant challenges to the organization: human visual estimates of how much tape has been applied are wildly inaccurate, and the accompanying images taken to perform the assessment are inconsistent. This makes it difficult to keep track of the aircraft's status and the evolution of this special kind of damage.  

AI to the rescue

Luckily the solution developed by KLM and Mainblades, allows to decrease the amount of extra work required because of the paint peeling issue. The inspection drone can easily and automatically navigate the hangar, taking pictures of the inspected area in a fraction of the time needed to achieve the same coverage with cherry-pickers, making it an excellent choice for the MRO. Once the inspection drone completes the data collection, the images are reviewed to determine the damage and amount of tape applied. This brings down the review time from ~3 hours to ~30 minutes, six times faster. 

A close-up of a metal surface

While reviewing pictures captured by the drone is already faster and safer than moving a human around the aircraft with heavy machinery, an additional speed up is achieved using Mainblades’ proprietary damage detection technology. The damage instances are predicted on the images and suggested to the reviewer who can quickly accept, discard, or modify them to be as accurate as possible. Lastly, since the automatic navigation of the drone uses 3D geometry information of the aircraft in the form of a 3D map, it can project where that finding is in the 3D world and accurately estimate its size. Sounds familiar? In the validation procedure carried out internally at KLM, it was found that the result of a drone inspection is 8.5 times more accurate than a manual inspection. 


learn more about implementing inspection automation

contact us