Abstract: Ship sub-assembly assembly is crucial in shipbuilding, but it faces many challenges. The shipbuilding industry is labor-intensive. While the number of new orders is increasing, it is facing low efficiency and "labor shortage". The manufacturing process of sub-assemblies is complex and susceptible to various abnormal factors, making production scheduling difficult. At the same time, the data in the manufacturing system is multi-source and heterogeneous, and it is difficult for the information system to obtain real-time comprehensive information, which affects the formulation of scheduling plans.Although cyber-physical fusion is crucial, existing research is not applicable to shipbuilding and has not solved the problem of data semantic unity. Research on production scheduling based on ontology also has limitations.Therefore, this article takes the assembly of ship sub-assembly as the research object and proposes an ontology-based semantic integration and dynamic scheduling decision architecture. Firstly, a method is proposed to construct a workshop scheduling ontology model, data model, and rule model, for establishing a unified data semantic system, and ensuring the rapid integration of multi-source heterogeneous data in the information domain. Secondly, a real-time data integration platform based on Flink CDC technology is established to ensure timely acquisition of physical domain assembly process data. Subsequently, an assembly process control system of ship sub-assembly is developed, which includes production monitoring, disturbance detection, and dynamic scheduling functions. Through real-time interaction of cyber domain and physical domain data, the manufacturing process of the ship sub-assembly is dynamically controlled. The effectiveness of the proposed system is verified by comparing before and after implementation.