Particle Swarm Optimization Based Interval Inversion of Direct Push Logging Data

Abstract

An interval inversion approach based on global optimization is suggested for the interpretation of direct-push logging data. To further increase the overdetermination ratio of the inverse problem, the result of factor analysis is incorporated into the inversion procedure. The direct-push logging dataset consists of natural gamma-ray intensity, electrical resistivity, bulk density, neutron-porosity and cone resistance logs. First, factor analysis is carried out to estimate the water content of unsaturated sediments distributed along the borehole. Then, interval inversion is done by utilizing the information of factor analysis on water content to estimate the remaining model parameters such as clay and sand content. Gas content of the studied formation is derived from the inversion results using the material balance equation. It is shown that the factor analysis assisted interval inversion procedure gives highly accurate estimation to the model parameters. As an added advantage of the hybrid method, the starting model dependence of the inversion procedure can be greatly reduced owing to the Particle Swarm Optimization (PSO) technique applied to solve the inverse problem.