The geological body has uncertainty and complexity characters, therefore the predesign scheme for construction has blindness. Field measurement and back analysis can help identify rock mass parameters in actual engineering field. The paper combines an intelligent optimization arithmetic-particle swarm optimization (PSO) and 3-D fast Lagrange numerical method to construct a parameters back analysis method for underground engineering. The PSO has global robust optimization property and overcomes the conventional optimization method problem of being limited in local optimization. The FISH language is used to develop the PSO arithmetic and is embedded in the business software- FLAC-3D. The method is used in Dalian Metro construction and it proves that this method has not only fast convergence speed but also high prediction. It offers a new idea to geotechnical parameters identification.