MULTI‐PARAMETER FULL WAVEFORM INVERSION FOR ACOUSTIC VTI MEDIA USING THE TRUNCATED NEWTON METHOD

In multi‐parameter full waveform inversion (FWI), the trade‐off between different parameter classes increases its nonlinearity severely. The magnitude orders of seismic velocities and anisotropic parameters are significantly different, which makes the inverse problem poorly conditioned. Judiciously incorporating the inverse Hessian operator into the FWI scheme can help to alleviate the negative impacts of these two problems in the inversion and improve the inversion accuracy. The truncated Newton (TN) method is one of the optimization methods which can better estimate the inverse Hessian operator. This paper studies the simultaneous inversion of two parameters for acoustic VTI media in the time domain using the TN method. Numerical experiments on different models show that in the context of simultaneous inversion of two parameters for acoustic VTI media, the TN method can account for more accurate inverse Hessian operator than the limited memory Broyden‐Fletcher‐Goldfarb‐Shanno (L‐BFGS) method so that it can better balance the updates of two different parameter classes, and reconstruct more accurate parameter models.