3D closed-loop surface-related multiple elimination based on GPU acceleration

Abstract

The Closed-Loop surface-related multiple elimination (CL-SRME) shares the common theoretical foundation with traditional surface-related multiple elimination (SRME). Nevertheless, it introduces an inversion-based approach to avoid the adaptive subtraction process in SRME, aiming to prevent the energy damage to the primaries that may occur when they interfere with multiples during multiple suppression. With the advancements of computing power, the seismic data for processing has evolved from 2D to 3D. However, traditional 2D algorithms are no longer sufficient to effectively suppress surface-related multiples in 3D data. Consequently, based on the theories of 3D SRME and 2D CL-SRME, the 3D CL-SRME algorithm is proposed in this study. Moreover, the implementation of the CL-SRME necessitates numerous matrix operations and frequent data conversions between the time domain and frequency domain, resulting in colossal computational costs. Therefore, a GPU acceleration strategy is introduced to address this challenge. Numerical examples of 3D seismic data demonstrate that 3D CL-SRME can provide higher accuracy of multiple suppression and wider adaptability to complex 3D cases. Simultaneously, the graphics processing unit (GPU) parallel computing can substantially enhance the computational efficiency. This study employs a novel approach that achieves significant improvements in performance and accuracy for surface-related multiple elimination tasks in 3D applications. The combination of its closed-loop approach and GPU acceleration renders it a valuable tool for 3D multiple suppression, enabling high-precision multiple suppression with less computational cost.