Nonlinear AVO Inversion Based on a Limited Memory Quasi-Newton Algorithm

Abstract

Summary The routine forward equation of AVO inversion became nonlinear when S-wave to P-wave velocity ratio is considered to be an unknown parameter. Then the nonlinear forward equation was used to establish the objective function for nonlinear AVO inversion based on the Bayesian theory. A limited memory quasi-Newton algorithm which is called L-BFGS algorithm was used to solve the objective function. This algorithm stores information from a limited number of previous iterations to form an inverse of the approximated Hessian, requiring moderate memory. The validity of the nonlinear AVO inversion strategy is proved by actual well logs.