Abstract: Through analyzing the influencing factors of water inrush from deep-well floor, a hierarchic analysis structure model is established. And then, the weights of the influencing factors are calculated and arranged in order by using analytic hierarchy process (AHP), through which the major influencing factors are decided. On this basis, the factors are normalized through setting up the membership function and membership grade, and a discrimination model for water inrush is constructed based on fuzzy neural network (FNN). By choosing appropriate network structure parameters and training samples, this model, with the grade of water inrush as the output, achieves the learning process. Finally, this model is verified with field examples, which shows that the result obtained by the model is basically in accordance with the engineering practice.