One thing is true in any industry – time is money. So when the Rolls-Royce Innovation Hub came up with the Intelligent Borescope that can reduce the time it takes to inspect an aircraft engine by 75% and could save up to £100m in inspection costs over five years, they knew they had created something that was not only cost effective but could potentially transform the future of aviation inspection procedures.
“Our new Intelligent Borescope capability brings together cutting-edge technology and AI in a way that has never been used in the aviation industry before,” says Adriano Pulisciano, Rolls-Royce Imaging and Computer Vision Specialist.
How does it work?
The tip of the Intelligent Borescope, which is about the size of a pen lid, is enabled with a scanner that can generate 3D colour images. It can view and scan whole objects as large as High Pressure Turbine blades. As it moves through the engine, the borescope captures images which are analysed by the AI App installed on the Borescope handset and then sent to the Rolls-Royce Cloud.
The Intelligent Borescope, powered with a menu driven inspection process, is directly connected to a Rhinestahl CTS Electronic Turning Tool that automatically and safely rotates the engine stage and positions each blade correctly. After each blade is successfully positioned by the turning tool, it communicates with the borescope to capture the image automatically.
The next step is to process the data in the App, which is achieved using AI technology similar to that used for facial recognition. The App maps the blade in the same way a face is mapped looking for key image features that represent inconsistencies or irregularities. This is the first time facial recognition AI technology has been used in this way in aerospace inspections and it required the team to design and build the AI network from scratch with representative data collected from across our fleet of engines.
Once the AI process is complete the operator has to review and approve the predictions. During the review process they can easily fine tune the predictions if needed. In the days before we had this App, processing this data for particular inspections took an operator 90 minutes – it now takes just five.