Development of a vision system enhanced by deep learning to support robotic laser cleaning

Robotic laser cleaning has become a promising technology in the field of surface cleaning due to its efficiency as well as green and environmentally friendly nature. This technology may replace chemical, mechanical, and ultrasonic cleaning methods thanks to its precision in removing contaminants from surfaces without causing damage, with simple automation and reduced costs. In this work, a vision system enhanced by deep learning is proposed with the aim to improve the performance of a robotic laser cleaning system designed to remove contaminants from dies used in composite laminate production in the aerospace industry. The proposed vision system, composed by a camera and a processing unit, allows to capture images from the die and process them using a Convolutional Neural Network, allowing for the classification of tool cleanliness levels with an accuracy of 95%. The introduction of the proposed system can boost production capacity, reduce costs and minimize operators’ efforts and environmental footprint, which may mark a significant innovation in the aerospace industry manufacturing.