Alternative Ways to Formalize Knowledge for Petrophysical Inversion

Abstract

SUMMARYRobust petrophysical inversion in addition to logging data (matrix Y) requires additional information about the desired physical properties (matrix X) and their interrelationships. This approach allows us to regularize the solution and avoid petrophysical unreality. The author earlier proposed to classify alternatives for specifying such interrelationships of physical properties using cluster analysis and regression. The paper uses a sample of petrophysical data published in open press to compare cluster-type algorithms of mixture type and cluster regression and to recommend improvements.