Abstract: The assembly of water joint is a critical step in the assembly of liquid cooling system for active phased array radar, requiring high stability in assembly quality and excellent sealing performance. To improve the stability and intelligence of water joint assembly, a digital twin system for a water joint assembly robot is studied and designed. Firstly, the functional requirements of the digital twin system for water joint assembly are analyzed, and a system architecture is proposed, including the physical entity layer, data layer, twin model layer, and application layer. Then, the digital twin system for the water joint assembly robot is developed through the acquisition and processing of multi-source heterogeneous data, the construction of digital twin model, and the development of functional modules. Finally, the application of the system is validated using the water joint assembly process of a specific product model. The results demonstrate that the developed digital twin system can achieve a comprehensive digital twin of the entire water joint assembly process, enabling real-time monitoring and traceability of the assembly status and screw tightening torque, as well as fault prediction. This system significantly improves the intelligence level of the water joint assembly process and ensures the stability of assembly quality.