The Inversion Method of Surface-Wave Frequency Dispersion Curve Based on Neural Network

In recent years, people began to make use of surface wave for the investigation of near-surface structure and had achieved some good results. This paper attempts to use neural network to the surface-wave frequency dispersion curve inversion. First, a geological model of four horizontal layers is designed, with each layer given different thickness, density, velocity parameters, and a finite-difference wave equation is applied for full-wave field forward modeling. Then the surface-wave records are extracted by singular value decomposition method, used to calculate frequency dispersion curves as neural network training samples. Finally, the trained neural network is used to inverse the formation parameters. The forecasting results show that neural network is an effective method to the formation parameters inversion.