The gradient preconditioning approach based on seismic wave energy can effectively avoid the huge storage consumption in the gradient preconditioning algorithms based on Hessian matrices in time-domain full waveform inversion (FWI), but the accuracy is affected by the energy of reflected waves when strong reflectors are present in velocity model. To address this problem, we propose a gradient preconditioning method, which scales the gradient based on the energy of the “approximated transmitted wavefield” simulated by the nonreflecting acoustic wave equation. The method does not require computing or storing the Hessian matrix or its inverse. Furthermore, it can effectively eliminate the effects caused by geometric diffusion and non-uniformity illumination on gradient. The results of model experiments confirm that the time-domain FWI using the gradient preconditioning based on transmitted waves energy can achieve higher inversion precision for high-velocity body and the deep strata below when compared with using the gradient preconditioning based on seismic waves energy.
|Original language||English (US)|
|Title of host publication||International Geophysical Conference, Qingdao, China, 17-20 April 2017|
|Publisher||Society of Exploration Geophysicists|
|State||Published - May 31 2017|
Bibliographical noteKAUST Repository Item: Exported on 2020-10-01
Acknowledgements: This work was supported by the National Natural Science Foundation of China (41574105).