The traffic flow is increasing with the development of urban modernization and it is becoming more and more important to ensure the smooth flow of traffic. Urban traffic network is a complex network with multi-node and intricate edges. Nodes or edges in complex networks may break down because of highly loaded when they are attacked on purpose. The breakdown can spread along the network and finally cause cascading failure. Therefore, it is necessary to study how cascading failure spread and how to stop its spread. We use the data from a city in the north of China to study cascading failure in urban traffic networks. Firstly, a cascading failure model is built based on an improved load-capacity model and we make it more suitable according to the features of traffic networks. We use this model to predict the distribution of traffic flow at next moment and the traffic flow after a small change in the network. Secondly, we consider the nodes’ importance degree by their betweenness and other features. We invent a method to delete proper nodes in order to control the spreading of cascading failure. The total transit time becomes shorter than before the node is deleted. Thirdly, we use SUMO, a simulation software, to simulate the real situation with the data. The results prove that proper deletion of nodes can increase network efficiency and reduce the impact of cascading failures. The simulation results verify the validity of the model to some extent. The research results have guiding significance for the actual situation.

CICTP