基于模糊Petri网的不确定工时柔性作业车间调度方法OA

In view of the lack of flexibility of traditional heuristic optimization algorithms when solving the FJSP(Flexible Job-Shop Scheduling Problem)with uncertain processing time of workpieces,this study proposes a DRL(Deep Reinforcement Learning)scheduling method based on FPN(Fuzzy Petri Net).A FPN model for discrete shops is constructed with the optimization objective of minimizing the total makespan,which simulates the dynamic characteristics and uncertainties of the shop environment.The uncertainty of processing times is handled by redesign-ing the fuzzy logic and combining it with an improved whitening weight function for interval grey numbers.An opti-mized DDQN(Double Deep Q-Network)algorithm is designed,which enhances the randomness of action exploration by adding noise to the network.An improved priority sampling method is developed to perform weighted sampling based on the importance of experience replay samples,improving learning efficiency and algorithm convergence speed.Experimental results show that compared with traditional heuristic scheduling algorithms,the proposed sched-uling method achieves a shorter total makespan,demonstrating its advancement and effectiveness in improving sched-uling efficiency and solving uncertainty problems.