The continuous mining in the coal mine has made the water quality characteristics of each aquifer become more complex and more similar, and it is difficult to establish a precise discriminating model by using classical mathematical methods. The BP neural network with nonlinear mapping function can overcome the above problems, but it still has the disadvantage of being easy to fall into local optimization and having slow convergence speed. We introduced the "premature" judgment mechanism, Tent chaos map and adaptive weight adjusting strategy into particle swarm optimization algorithm, establishing a water inrush discrimination model on the basis of adaptive chaos particle swarm optimization algorithm and BP(ACPSO-BP) neural network. The application results show that compared with the BP neural network model and the model based on standard particle swarm algorithm and BP(SPSO-BP) neural network, the ACPSO-BP neural network model has faster convergence speed, higher accuracy and stronger generalization.