ML-supported, flexible screw disassembly framework for the remanufacturing process of complex capital goods

Remanufacturing with efficient dismantling is important for sustainable manufacturing, but condition variability of end-of-life products complicates automated, non-destructive disassembly. Exposure to challenging environmental conditions often results in seized screw connections leading to damage and loss of high-quality fasteners. This work presents a framework for complex capital goods that combines operational data with a manually operated, piezo-actuated hand tool that generates controlled low-frequency vibrations to reduce loosening torque. A data-driven, machine learning-based process selects the optimal vibration parameters from usage data. Tests on specimens and real engines have demonstrated torque reductions of up to 20 %, which decreases the need for rework.