Solving Disassembly Line Balancing Problems via Q-learning Based Iterative Algorithm With Minimizing Cycle Time

This work investigates type II disassembly line balancing problems. The objective is to minimize the cycle time with a fixed workstation number. First, a mathematical model is built to describe the concerned problems. Interference time between tasks is represented by an interference time matrix. Second, seven local search operators are developed according to problem-specific feature and knowledge. Q-learning is employed to select appropriate local search operator during algorithm’s iterative process. Finally, 11 instances are solved. Comparisons and discussions show that the proposed iterative algorithm combining with Q-learning based local search can solve the disassembly line balancing problems well.