Seismic Tomography of the Velocity Structure of the Upper Mantle of the Earth’s Crust Based on Artificial Intelligence Technology

The velocity at the top of the upper mantle of the Earth’s crust is an important parameter to study the evolution of the Earth’s crust. The formation and evolution of surface tectonic units are closely related to the velocity structure at the top of the upper mantle. The propagation time of seismic waves between earthquakes and stations in modern seismology has high accuracy and reliability, including the study of the Earth’s ring structure and the source of deep driving force of plate tectonics are mainly from the interpretation of velocity structure, using the time information to extract velocity information is the most important method in geophysical methods, which has been widely applied to obtain the velocity structure of the crustal mantle. Seismic tomography is one of the most effective means to study the velocity structure of the Earth’s interior. The results obtained by manual extrapolation and distribution assumptions are sometimes unsatisfactory due to the difficulty of knowledge acquisition and the limitations of experts’ knowledge of multidimensional data. With the development of artificial intelligence technology, the interpretation accuracy and credibility of seismic laminar imaging can be improved by using the self-organized learning ability possessed by artificial neural networks and their powerful classification computational power. Therefore, the method studied in this paper is a seismic walk-time laminar imaging method that uses artificial intelligence technology and seismic-related data to invert the velocity structure of the upper mantle of the Earth’s crust. The method can be divided into four parts: model parameterization, calculation of travel time and path, inversion calculation and performance evaluation. The first part is to divide the model space of the study area into blocks or grid nodes (model parameterization), calculate the path and travel time of seismic wave rays from the source to the station, and then build a neural network model based on the difference between the theoretical travel time and the observed travel time of seismic waves, i.e., travel time residual, and finally obtain the seismic laminar imaging of the study area.