Fine-Grained Visual Categorization of Fasteners in Overhaul Processes

Commercial aircraft engines must be overhauled approximately every six years, during which hundreds of different parts must be disassembled, checked, and then reassembled. This includes undoing up to thousands of fasteners, cleaning, checking, refitting, and tightening them. Prior to refitting the fasteners, they must be classified and packaged. In this paper, we describe a system for classifying fasteners automatically, by use of computer vision and machine learning. Using the proposed system, we created sample datasets and performed a fine-grained visual categorization of the fasteners. Our trained model classifies 20 bolts and 14 washers with an accuracy of 99.4%. Our work is the first step towards an automated fastener classification system in overhaul processes.