Abstract: This paper firstly used FLAC3D for simulating the surface subsidence of the old goaf of the Hengshan gypsum mine where has been collapsed. Secondly, simulating the surface subsidence of the old goaf with different structural parameters has been used and analyzed the factors which is influencing the surface subsidence. Then, the simulating results were used for establishing the ANFIS (adaptive neuro-fuzzy inference system) model for predicting the surface subsidence of the old goaf and the fitting and predicting capabilities were analyzed. Finally, the ANFIS model was used for predicting the surface subsidence of the old goaf of a gypsum mine in Hengshan and the practical surface subsidence was measured. The results show that, the accuracies of the simulating results by FLAC3D amount to 97%, and with the mechanical parameters fixed, room span, interval pillar span and stope height are the key parameters influencing the surface subsidence of the old goaf. The accuracies of the fitting and predicting results by the established ANFIS model amount to 96% and 97%, respectively, and the predicted results of the surface subsidence of the old goaf of a gypsum mine by the established ANFIS model and the measured results are in good agreement. The interval pillar span and room span are suitable to 4 m and 10m for the mining stopes of the gypsum mine in Hengshan.

Key words: gypsum mine, surface subsidence, structural parameters of mining stopes, FLAC3D, ANFIS model