Abstract:

The presence of multiple waves will cause serious distortion of seismic data imaging, increase the interference of the target layer, and thus have an adverse impact on the accurate interpretation of geological data and the development planning of oil and gas reservoirs. Although the parabolic Radon transform based on L1/2 norm constraint shows excellent results in multiple suppression, the traditional fast iterative shrinkage-thresholding algorithm (FISTA) is difficult to meet the requirements in terms of computational efficiency when facing large-scale seismic data. To this end, this paper proposes an improved greed-fast iterative shrinkage-thresholding algorithm (G-FISTA). This algorithm achieves a significant improvement in the inversion calculation efficiency by optimizing the iterative steps and combining convergence conditions with a restart mechanism. The numerical test results show that compared with  FISTA, the convergence speed of G-FISTA is significantly improved, and it shows good multiple suppression effects in both synthetic data and actual seismic data. Compared with the traditional algorithm that iterates ten times to reach the optimal solution, G-FISTA  only needs four iterations, nearly doubling the iteration efficiency.

Key words: multiple waves, attenuation, computational efficiency, Radon, G-FISTA