A novel deep learning-based automatic damage detection and localization method for remanufacturing/repair

Abstract

Remanufacturing has been considered as an eco-industry, demonstrating environmental and economic benefits. Damage feature inspection is a critical and step in remanufacturing, which establishes the connection between used part and process planning. However, the current inspection method for remanufacturing heavily relies on manual operations. In this study, a deep learning-based damage recognition and spatial localization method is developed. The damage recognition method is based on a Mask-RCNN model to output damage type, 2D damage segments. By mapping the 2D pixel coordinates to the 3D global coordinate system, the spatial coordinate of damage is calculated. With identifying and positioning damages, further automatic repairing/remanufacturing processes can be operated based on these results.