The Encoded Multisource Full Waveform Inversion And Real Date Test

Full waveform inversion(FWI)has been a hotspot in geophysical exploration because it can make full use of the observed seismic data.We can obtain a high resolution result by FWI,which has incomparable advantages compared with other inversion methods for the inversion of subsurface structures in complex areas.In FWI,an objective function consisting of the difference between the real field data and the synthetic data is defined.The parameters of subsurface to be inverted are obtained by calculating the minimum of the objective function through multiple iterations.We discuss the research status of FWI firstly and then review the basic method of forward simulation of wave equation.Based on the 1st-order velocity-stress elastic wave equation,we calculate the back-propagated wavefield and the gradient of the objective function with respect to the model parameters according to the Lagrange adjoint theory.We update the model iteratively by the conjugate gradient method,apply the convolutional objective to avoid extracting accurate wavelet and design a multi-scale inversion strategy to improve the convergence effect of inversion.Then,we deal with the huge computation of FWI and improve the computational efficiency by the source-encoding method.In order to overcome the problem of cross-talk noise and acquisition geometry mismatch in conventional source-encoding method,we utilize the orthogonality of trigonometric functions.We use harmonic sine functions of different frequencies as sources,encode them into a super shot and then separate the wavefields with different frequencies by the phase sensitive detection method.This can theoretically remove the cross-talk noise.Based on this,we apply a frequency selection filtering to extract the single shot seismic data from the encoded synthetic data.We can improve the efficiency of inversion while ensuring the precision of inversion.Finally,we use the real data to test the full waveform inversion.