RAYLEIGH WAVE DISPERSION CURVE INVERSION COMBINING WITH GA AND DSL

摘要

Rayleigh wave dispersion curve inversion is a multi-parameter highly non-linear iterative optimization process. The conventional single linear or non-linear inversion method has some limitations for the complex seismic geologic conditions, which can lead to more prominent multi-solution problem. But the defects both of the methods can be supplemented by the advantages of each other. In order to further improve the inversion accuracy, we proposed a joint inversion method via complementing and nesting the linear (damping least squares) and non-linear (genetic algorithm) methods. Firstly, the genetic algorithm (GA) is utilized based on the loose constraints of prior geological information to lock in the target near the global optimal solution. Then, use the damping least squares (DLS) method to achieve higher precision of Rayleigh wave dispersion curve inversion. The effectiveness of the method has been verified by a typical layered model. And we use the method to further process actual seismic data. Results show that the method not only absorbs the advantages of GA with global optimization and strong adaptability, but also inherits the advantages of DLS with fast convergence and stable inversion. And better results are achieved in suppressing multi-solution, getting rid of the initial model highly dependent, and improving inversion accuracy.