Machine learning-based system to automate visual inspection in aerospace engine manufacturing

Abstract

Presented is a system to digitalize, assist and automate visual inspection of aerospace component surfaces. Industrial robotic and camera equipment are applied for image acquisition, while deep learning models are used for automated defect detection. Operator interaction for evaluation of indications is handled in a user interface developed for this purpose. Inspection data is stored in a database, allowing to trace back results and continuously improve the system. The system is designed as a platform for different sensors and different inspection processes. The conducted development contributes to increase efficiency, effectivity and traceability of inspection processes in the aerospace industry.