The analysis of robustness, an important performance index of urban transit networks, helps to well evaluate and optimize the measures of urban transit systems. In this paper, first, three models of urban transit systems respectively for the original transit network, the transit station network and the transit line network are established based on the bipartite graph model, and a robustness index of the topology structure of urban transit networks is de- fined. Then, a fast algorithm is proposed for the robustness analysis of large-scale networks. Finally, the robustness of the three established models is analyzed, with Beijing transit system under random and intentional attacks as an example. The results indicate that, compared with the random network, the urban transit network is not sensitive to the random attack, but somewhat sensitive to degree-based and betweenness-based attack.