报告人：廖文远（University of Calgary）

报告题目：A computational method for solving full-waveform inversion problem through PDE-constrained optimization.

报告摘要：Full-waveform inversion (FWI) is a model-based data-fitting technique that has been widely used to estimate the medium property in Geophysics community for the purpose of hydrocarbon exploration.  Given the large size of observational seismic data, and the iterative nature of the numerical procedure, FWI is computationally intensive, moreover, such inverse problem is ill-posed in general.  Various numerical and theoretical strategies can be used to improve the well-posedness. In this work we formulate the FWI of crosswell seismic data as a PDE-constrained optimization problem, which is solved by the popular adjoint state method. The objective functional to be minimized is defined as the square norm of the misfit between observational data and modeled data based on an initial model, while the model parameter to be recovered is the acoustic velocity, which appears as the coefficient in the seismic wave equation. The issues that we attempt to address here include the creation of a reliable initial model, gradient calculation method for instance the automatic differentiation tools and gradient-based optimization algorithms. Some preliminary numerical results are presented to demonstrate the effectiveness and efficiency of the proposed numerical approach.