A time domain modeling and inversion technique and its applications to electromagnetic subsurface sensing.

摘要

Forward modeling and data inversion of subsurface-sensing systems and well-logging tools are very useful not only because they help us to understand tool performances in unknown circumstances, diagnose malfunctions, and interpret measured data, but also guide new tool designs. In this dissertation, numerical techniques for forward modeling and data inversion in the time domain are developed for the purpose of analysis, design and data interpretation of some subsurface-sensing and well-logging tools. The forward modeling is based on the transmission line matrix (TLM) method and is further developed to simulate subsurface sources. The modeling technique is verified and applied to new tool designs by simulating a cross-borehole electromagnetic tomography system and measurement-while-drilling (MWD) resistivity tools. A new type of MWD tool with electric dipoles is proposed and performance is discussed. Its ability to detect horizontal and concentric formation-bed boundaries is found superior to the MWD tools with magnetic dipoles. For the measured data inversion, an iterative algorithm is developed to reconstruct the image of the formation conductivity surrounding a borehole using the time-domain data. Since the inversion algorithm proceeds iteratively and the part of the formation involved in each iteration marches in space step by step, no optimization procedure is necessary. Problems caused by optimization process, such as finding the solution of a large-scale system equations and computation of Jacobian matrix numerically, are avoided. This method is especially useful in cases where the analytic gradient is not available. The inversion algorithm is tested in formations having both one- and two-dimensional conductivity variations. Reconstructed results show that the proposed inversion method is of satisfactory resolution and investigation depth.