摘要: The displacements of strata at different heights of an undersea gold mine during the mining process are monitored. The time series of stratum displacement are reconstructed in phase space. The changing laws of distance between two phase points for the displacement of strata at different heights in the phase space are revealed using the chaos theory. A prediction model for the evolution laws of phase space distance of stratum displacement is established based on the neural network, by which the stratum displacement of undersea mining in Xinli mining area is predicted. Then the security early warning system of strata displacement for the undersea mining is established. A neural network is trained through the combination of gradient descent method and chaos optimization method. The neural network model can achieve the merit of rapid training. Meanwhile, the defect of local minimum is avoided, and the calculation precision of the model is improved. The results show that the strata at different heights have different chaotic behaviors. After the reconstruction of phase space, subtle features of strata displacement change are enlarged, and the inherent law of strata is adequately demonstrated, which is the basis of the security warning system of the undersea mining.