Model-based Systems Engineering (MBSE) has been widely used in the development of complex systems. The system architectures, organizations, research and development processes using MBSE to design complex systems can be seen as a System of Systems (SoS), which has high complexity and hard to manage. The concept of digital thread is proposed to integrate all the models and data in the SoS. However, lack of cognition ability makes it hard to connect the models and data with human, processes and things in the SoS, which reduces the efficiency of complex system development. In this paper, a new concept named Cognitive Thread is first proposed as digital thread with augmented semantic capabilities for identifying the information of the SoS. Then a cognitive thread construction approach based on Open Services for Lifecycle Collaboration (OSLC) specification and knowledge graphs is proposed to support decision-making and management in the SoS. Finally, the feasibility of the proposed approach is verified through a case study of the advanced driver-assistance system development.