Optimization of Dynamic Scheduling in Additive Manufacturing with Deep Reinforcement Learning

Additive manufacturing (AM), also known as 3D printing, offers innovative solutions for mass customized production. This paper addresses the dynamic batch processing scheduling problem in AM, which considers release time and machine eligibility constraints, with the objective of minimizing total tardiness. To achieve this, a dueling deep Q network (DDQN) algorithm is proposed to generate an adaptive rule for batch formation and scheduling. In the proposed approach, the problem is formulated as a Markov decision process (MDP) by defining the state space, action space, and reward function. Three-dimensional features, encompassing dynamic orders, AM machines, and delays, are extracted to represent the production status at a rescheduling point. Additionally, five novel composite scheduling rules are introduced for selection when an AM machine completes processing or a new order arrives. Computational results demonstrate the effectiveness of the proposed DDQN algorithm in addressing the problem. A comparison is made against single proposed scheduling rules, randomly selected rules. The proposed DDQN algorithm showcases its capability to tackle the dynamic batch processing scheduling problem, offering improved performance.