Disassembly operations are fundamental in closed-loop manufacturing to recover components from returned products. Planning and optimization are obstructed by uncertainties in these operations, including unpredictable product conditions and incoming flow. Simulation is commonly used to support disassembly, but most tools lack flexibility and do not allow modular, scalable scenario definitions that reflect these uncertainties. This study proposes a modular methodology combining discrete-event simulation and machine learning. Disassembly scenarios are modeled with design parameters, and uncertainties (e.g., return quality) are applied using statistical distributions. This integration enables fast predictions of performance metrics (e.g., production time) and supports user interaction for optimizing decisions. A case study on a mini drone disassembly was conducted to demonstrate the performance of the proposed methodology. The case study demonstrates that the optimized scenario achieves a 30% reduction in operational time compared to the baseline. Applying the proposed methodology can help to make disassembly operations more efficient.
Oqueña et al. (Thu,) studied this question.