4d Printing-Enabled Circular Economy : Disassembly Sequence Planning Using Reinforcement Learning

Circular economy has been a rapidly growing interest in the manufacturing industry. The concept of 4D printing-enabled active disassemblyis touted as a non-destructive approach with less to no human intervention, which allows multiple elements to be disassembled simultaneously with fewer damages, labor fees, and energy. Disassembly sequences, methods, and end-of-life (EOL) options can be optimized to achieve higher recovery potential. This problem is remarkably underexplored in the current literature. Considering various uncertainties, a multi-objective mathematical model is established to examine the economic feasibility of adopting 4D printing for active disassembly and to bridge this gap. A modified deep reinforcement learning (DRL) algorithm is proposed to solve general disassembly sequence planning problems efficiently. In addition, a new grid-world product representation method is proposed to encode detailed product information. The effectiveness of the proposed algorithm is examined by comparing it with six state-of-art methods in real-world cases, and the influences of different encoders on the DRL are studied. Lastly, sensitivity analyses are conducted to investigate the influences of labor cost and manual disassembly damages on recovered profit