Towards Early Damage Detection during the Disassembly of Threaded Fasteners using Machine Learning

Abstract

At regular intervals, aircraft and their components, such as engines, are inspected following specified maintenance, repair and overhaul (MRO) guidelines. The modular design of the engines supports complete or partial disassembly for inspection of all relevant parts. The often used bolted joints are significantly altered by the harsh environmental conditions, severely increasing the disassembly effort usually carried out manually. Using tools like a ratchet or rotary impact wrenches, workers apply torque on the screw head to loosen the screw. If the loosening torque exceeds the material limits, screw heads are torn off, leaving the shaft in the base thread. This article presents a strategy to prevent disassembly damages through precise monitoring of the loosening torques and angle of rotation. Based on this data, a machine learning algorithm detects and predicts potential breakage, to allow adapted disassembly strategies to prevent complex rework. The algorithm will classify potential damage into different categories. Preliminary testing proved the applicability of machine learning toward aircraft disassembly.