| The purpose of this dissertation is to develop a method to invert induction and measuring-while-drilling (MWD) logs. The constructed formation model is two-dimensional, which includes a vertical borehole and some invasion zones. The bed boundaries are assumed to be horizontal and the inhomogeneous medium to be axially symmetric.; There are two essential components for inversion: fast forward modeling codes and convergent inversion algorithms. A set of time-efficient forward modeling codes based on a numerical mode matching method is employed to speed up the inversion procedure. Two convergent inversion algorithms are employed and completed in this research: the maximum entropy method (often referred to as MEM) and the Marquardt-Levenberg method (MLM). These two stable algorithms are useful for inversion using an iteration procedure, that is, to minimize the difference between the field logs and the simulated logs step by step.; The reconstructed profile for MEM is a 2-D formation model that includes a borehole, two fixed-radius invasion zones, and one uninvaded zone. The inverted formation for MLM is a profile that has a borehole, one invasion zone with an adjustable radius, and one virgin zone. The robustness and CPU time of the results produced by these two algorithms have been compared, and the MLM algorithm has been found to be more suitable than the MEM algorithm for inversion.; The codes developed by using these inversion methods are capable of inverting a single log as well as multiple logs. Some computer-simulated as well as field-log inversion cases are given. |
