Abstract: Effective cognition of data is the key factor to realize intelligent manufacturing. Aiming at a large number of multi-source and heterogeneous manufacturing data generated by industrial internet of things system, we propose a knowledge graph cognitive manufacturing model for industrial internet of things (IIoT-KGC). It built a knowledge graph model by cognitive-drive agent and proposes knowledge reasoning method based on deep reinforcement learning to realize the effective cognition of manufacturing resources in the industrial internet of things. Taking the personalized product order response of flexible workshop as an example, the experiment shows that under the dynamic demand change, IIoT-KGC has large positive sample rate as well as has better lathe utilization and real-time interaction ability than manual method and rule method, which provides technical support for intelligent manufacturing of industrial Internet of things.