Abstract：Based on orthogonal design，difference method，neural networks and genetic algorithms，the intelligent inversion is performed for the excavation displacement of underground powerhouse of Langyashan pumped- storage power station. Referring to the geological data，the different parameters and levels are chosen for rock and dike according to orthogonal design. The related analytical samples for neural networks are given by FLAC3D calculation. Through training the neural network with samples，the optimized inversion parameters are obtained by seeking for the optimum of genetic algorithms. The obtained parameters are used to calculate by FLAC3D and the results agree well with the in-situ measurement data.
