Xuqing Wu

This book presents a comprehensive introduction to well logging and the inverse problem. It explores challenges such as conventional data processing methods’ inability to handle local minima issues, and presents the explanations in an easy-to-follow way. The book describes statistical data interpretation by introducing the fundamentals behind the approach, as well as a range of sampling methods. In each chapter, a specific method is comprehensively introduced, together with representative examples. The book begins with basic information on well logging and logging while drilling, as well as a definition of the inverse problem. It then moves on to discuss the fundamentals of statistical inverse methods, Bayesian inference, and a new sampling method that can be used to supplement it, the hybrid Monte Carlo method. The book then addresses a specific problem in the inversion of downhole logging data, and the interpretation of earth model complexity, before concluding with a meta-technique called the tempering method, which serves as a supplement to statistical sampling methods. Given its scope, the book offers a valuable reference guide for drilling engineers, well logging tool physicists, and geoscientists, as well as students in the areas of petroleum engineering and electrical engineering.

Seismic imaging is a key component of subsurface exploration, and it depends on a high-quality seismic data acquisition system with effective seismic processing algorithms. Seismic data quality concerns various factors such as acquisition design, environmental constraints, sampling resolution, and noises. The focus of this book is to investigate efficient seismic data representation and signal enhancement solutions by leveraging the powerful feature engineering capability of deep learning. The book delves into seismic data representation and enhancement issues, ranging from seismic acquisition design to subsequent quality improvement and compression technologies. Given the challenges of obtaining suitable labeled training datasets for seismic data processing problems, we concentrate on exploring deep learning approaches that eliminate the need for labels. We combined novel deep learning techniques with conventional seismic data processing methods, and construct networks and frameworks tailored for seismic data processing. The editors and authors of this book come from both academia and industry with hands-on experiences in seismic data processing and imaging.