主分量分析的類神經網路於震測圖形分析與消除重複反射

Abstract

在本論文中將利用主分量分析類神經網路來求震測資料的相關矩陣之主要分量。主分量分析的類神經網路可用來抽取震測反射層及有一致性的相鄰traces。在此將用類神經網路來做重複反射的消除。Huang 針對主分量分析提出一個理論，此理論為在現有的特徵向量方向上加上額外的點，可加強此特徵向量，此理論將應用於斷層分析及實際的震測圖上。 對於重複反射的消除，先對震測資料作速度分析及動態修正，使得重複反射層顯現出一致的水平性質，然後便可用主分量分析類神經網路將重複反射分離出來。重複此程序直到所有的重複反射都被消除，即可得到不含重複反射的震測資料，以利往後對震測資料作進一步的處理及解釋。

The neural network using an adaptive learning algorithm (ALA) is adopted to find the principal eigenvectors of a covariance matrix in different kinds of seismograms. Principal components analysis (PCA) using the neural net of ALA allows one to extract information of seismic reflection layers and uniform neighboring traces. The neural network is applied to the removal of multiples. And for PCA, Huang has proposed a theorem that adding extra point along the direction of the existing eigenvector can enhance that eigenvector. This theorem is applied to the fault analysis and a real seismogram at Mississippi Canyon. About multiple removal, velocity analysis and normal moveout correction will be applied to the seismogram first, so that the multiple reflection will show uniform horizontal property. Therefore, the multiple can be extracted from the seismogram by PCA using neural network. The procedure will be repeated to remove another multiple. In the end, the seismic data contain the reflection without multiples, and can be used to the further seismic data processing and interpretation.