Memoryless quasi-Newton (MLQN) method for 2D acoustic full waveform inversion

Abstract

Full waveform inversion (FWI) is an efficient way to solve parameter reconstruction problems, such as velocity, density, and viscosity coefficient. In this study, we apply a memoryless quasi-Newton (MLQN) method in FWI to invert velocity from surface seismic data for the first time. This method can attain acceptable results with low computational cost and small memory storage requirements. To ensure that the inverted velocity is maintained between the lower and upper boundaries of the velocity model, a nonlinear transformation is added to velocity as a priori information. To test the efficiency of the MLQN method in FWI, two synthetic models, a modified Marmousi model and a modified overthrust model, are examined from the surface seismic data with and without white Gaussian noise. For comparison, the conjugate gradient (CG) method is carried out for the same velocity models with the same parameters. We compare the inverted velocities by the two methods based on the aspects of memory storage requirements, computation time for each iteration, and error. By keeping the memory storage requirements and computation time in each iteration similar, the reconstructed velocity models obtained using the MLQN method are closer to the true velocity models than those obtained using the CG method. Our numerical tests show that the MLQN method is feasible and reliable in FWI.