For complex metro systems to operate efficiently, effective cooperation among teams is needed. Many accidents happen due to control failure caused by team errors during cooperation. Improving the efficiency of teamwork not only focuses on the team’s organizational structure and teamwork procedure, but also stresses each workers’ state during cooperation, because personal errors or interactions errors may lead to the collaboration failure. Combining the theory of human error and the complex network theory, this paper proposes a method to identify team errors and explore the mechanisms of team error during teamwork in the metro systems. Team cooperation is decomposed into each team member’s information process and interactions among human, machines and the environment, which are connected by a collection of information flows in the task. The TICN (Teamwork Information Complex Network) is established, in which information state of individual and interactions within agents are represented by nodes, and information flows are represented by directed edges. Then, vulnerable information states causing teamwork failure are identified by analyzing the degree centrality and betweenness centrality of the complex network. In the end, the model is applied to a specific emergency. The result shows that team members need enhanced skill-based level cognitive ability, the station supervisor needs a thorough understanding of the management system and rules, the decision-maker has to make and communicate decisions concisely, and the communication equipment should be adjusted and designed according to ergonomic requirements. Based on the results, some suggestions were put forward to improve the team performance in the metro system.