The manufacturing process of traditional Chinese medicine is subject to material fluctuation and other uncertain factors which usually cause non-optimal state and inconsistent product quality. Therefore, it is necessary to design and collect the quality-rela-ted physical parameters, process parameters, and equipment parameters in the whole manufacturing process of traditional Chinese medicine for digitization and modeling of the process. In this paper, a method for non-optimal state identification and self-recovering regulation was developed for active quality control in the manufacturing process of traditional Chinese medicine. Moreover, taking vacuum belt drying process as an example, a DQN algorithm-based intelligent decision model was established and verified and the implementation process was also discussed and studied. Thus, the process parameters-based self-optimization strategy discovery and path planning of optimal process control were rea-lized in this study. The results showed that the deep reinforcement learning-based artificial intelligence technology was helpful to improve the product quality consistency, reduce production cost, and increase benefit.
Keywords: deep reinforcement learning; intelligent manufacturing; process optimization; self-decision; traditional Chinese medicine manufacturing.