Study On Method Of Internal Multiple Elimination

Currently, seismic interpretation focuses on the reservoir property instead of geologicstructure, and its request for quality of seismic data is higher and higher. But the existence ofinternal multiple hinders the improvement of process result because of its similarity toprimary. Up to now, there are a few methods of internal-multiple-elimination (IME), but theycan’t give ideal results when used to real data, so it is necessary to continue research on suchproblems. In this paper, I have done a further research on data-driven internal multipleelimination method based on previous theories, and prospect to give a valuable suggestionwhen such method applied on real data.Firstly, concept and basic characteristics of internal multiples have been discussed in thepaper. Two geologic models, layers with uncomfortable surface and layers with geologicalabnormity body, were summarized by theoretical analysis and forward modeling, which cangenerate some stronger internal multiples. I have also expanded upon the concept of layerrelated internal multiple as upper layer related internal multiple (ULRIM) and lower layerrelated internal multiple (LLRIM) whereby ULRIM was mostly studied in past studies.Through the study, it is proved that ULRIM can’t be overlooked when comes to internalmultiple.As the same to other wave-equation-based multiple elimination methods, IME discussedin this paper need two steps to complete—multiple prediction and subtraction. The two partsare divided into two chapters in this paper, whereby multiple predictions is the core content. Ihave re-derived formulas of upper layer related internal multiple prediction, and give the formfor lower layer related internal multiple. I implemented the time-domain prediction andintroduced the procedure by sample examples. Then, I realized frequency-domain algorithmby comparing fake-back process for surface related multiples. Experimental results show thatfrequency-domain algorithm has the same effect and higher computational efficiency. As for multiple subtraction method, the paper involves two most common methods—least-square based method and constrained cross-equalization method. Subtraction algorithmswere introduced and criteria for selection of parameters were discussed.Finally, the paper shows the application of involved algorithm on two complex mediummodel dataset in allusion to ULRIM and LLRIM. In addition, the effect of stochastic noise forinternal multiple elimination process was discussed, such work can provide reference framefor real data application.