Abstract: Identification of critical nodes in a network is one of key issues in complex network analysis. The input-output (IO) model in economics can be used to evaluate the node centrality in open flow networks, but it cannot be directly applied to closed flow networks which lack of external inflow and outflow. This paper introduces a virtual external node to convert the closed flow network into an open flow network, which makes the standard IO analysis method can be used to identify critical nodes in closed flow networks. This virtual external IO analysis method is then applied to the identification of critical nodes in Chinese Railway Network and the World Food and Agriculture Trade Network. The proposed analysis method provides an alternative approach for evaluating node centrality in closed flow networks.