The factors that influence the piping occurring in embankment are uncertain and random, and in the process of piping they turn out to be nonlinear behavior. RBF neural network theory was used to establish a new model in order to predict the piping of embankment. This paper discussed the establishment of predicting model and process of application in detail. Comparing with BP neural network, the application in 23 cases of embankment piping shows that RBF network be more precise, more quickly and faster.