Multifactory Disassembly Process Optimization Considering Worker Posture

The escalating consumption and disposal of electronic products have spurred a pressing demand for environmental conservation. Traditional disassembly factories encounter challenges when handling discarded products from various locations, including high costs and limited flexibility. This study addresses a multifactory disassembly process optimization problem, taking into account worker posture and the selection of disassembly line types. Subsequently, a mathematical model to maximize profit is built. The reinforcement learning algorithm, Categorical deep Q network (DQN), is utilized to find optimal solutions. Experimental results are compared with those from CPLEX to validate the precision and viability of the proposed model. Furthermore, we compare the proposed solution with various reinforcement learning algorithms, including DQN, proximal policy optimization, and Advantage Actor–Critic. The effectiveness of the proposed model and algorithm is verified by experiments on several cases with different complexity scales.