Big Data Driven Intelligent Production Control of Discrete Manufacturing Process

Abstract

Under intelligent manufacturing era, there is pressing demands from discrete manufacturing enterprise to utilize big data (BD) technologies for enhancing the level of production management and control (PM&C). The BD driven intelligent PM&C in discrete manufacturing process is studied. Based on the determination of characteristics and demands for PM&C, the architecture of BD driven PM&C is firstly constructed, which the main flow is "collection-processing-analysis-service" of manufacturing BD. Based on the closed-loop mechanism "progress prediction-bottleneck discovery-anomaly tracing-decision making" for PM&C, the key technologies are respectively proposed, which are: "A stacked sparse auto-encoder model for production progress prediction", "The parallel gated recurrent units model for shifting bottleneck discovery", "The density peak-weighted fuzzy C-means method for anomaly tracing" and "The multi-agents reinforcement learning for production decision-making". Finally, an aircraft discrete manufacturing workshop is selected as the application scenario to verify the developed prototype system.