Multiple Attenuation With SVD Technology

The multiple is a kind of coherent noise which always exists in seismic records, especially in marine dataset. How to eliminate or attenuate variety of multiples efficiently and reserve the characteristics of the primaries is a critical task in seismic data processing.By this time, methods that suppress multiples can be classified into two broad categories: (1) the wave equation methods that predict and then subtract multiples from the input data and (2) the filtering methods that seek to exploit a property that differentiates primary from multiple in kinematics. Using wave equation methods often means too much time-consuming and high-quality seismic data requirement.Therefore, the latter approaches with relatively simpler computations are widely used in industry.The conventional filtering methods are usually involved first in stacking velocity estimation of primary or multiple reflections, and then remove multiples on CMP gathers. Obviously, the accuracy of velocity analysis affects much to the performance of multiple suppressing. Besides, this kind of methods, which are based upon the normal moveout correction, are limited to the model of horizontal layer intrinsically. In this paper, we propose a new eliminating multiple approach, which combines the event realigning with the singular value decomposition (SVD) technology.In the dissertation we illustrate in detail how to utilize the property of SVD technology to attenuate multiples. The process of our approach to suppress multiples is as follows: (1) track reverberations via accurate ray tracing and divide the shot gather into different reasonable time windows according to the traced time-distance curves of reverberations. Within each time window, the multiple events are aligned to be horizontally flat. (2) filter out reverberations with SVD technology, which is very effective to eliminate the flat events; (3) re-align the filtered data back to the original shot gather. Remove all reverberations tracked by ray tracing, and then, we can complete the multiple attenuation. Our approach takes the mechanism of multiples into account adequately and requires no model. Further more, we can remove the multiples directly in shot gather without analyzing the velocity spectrum jumbled by multiples.The processing of synthetic and real examples clearly demonstrates the validity of the technology presented in this thesis.