Some applications of principal component analysis: Well-to-well correlation, Zonation
István Elek · Repository of the Academy's Library (Library of the Hungarian Academy of Sciences) · 1988
Principal Component Analysis (PCA) is a multivariate statistical technique often used successfully in various scientific disciplines. This paper aims to show the mathematical principles of PCA and introduce two log analysis applications based on the technique: well-to-well correlation and zonation. Traditionally well-to-well correlation has been performed using only one log, often a resistivity log or gamma ray log. A better, though usually slower, correlation often can be obtained by using all available wireline logs. This paper describes a method of computing the first principal component, a method which should make correlation easier, more efficient and accurate. The first principal component should contain the largest common part of variances of the input logs. (The Principle Component Analysis technique is described in the Appendix at the end of this paper.) The second part of this paper deals with a zonation technique based on the first principal component. The technique computes not only boundaries but characteristic values of log responses within a given layer. This computation is based on the ideal case that rock properties are constant within a layer and change suddenly at a layer boundary.