Development of a 3D Seismic Imaging System for Subsurface Fault Detection

Thesis Abstract

The accurate detection and characterization of subsurface faults remain critical challenges in geophysical exploration, given their implications for earthquake risk assessment, hydrocarbon exploration, and civil engineering development. Traditional seismic imaging techniques often suffer from limitations in resolution and accuracy when delineating fault structures, especially in complex geological settings. This study aims to develop a novel three-dimensional (3D) seismic imaging system optimized for subsurface fault detection, integrating advanced data processing algorithms with existing seismic data acquisition techniques to enhance fault delineation precision. The specific objectives include 1) designing an innovative 3D seismic data processing framework; 2) implementing the framework with real seismic datasets; 3) evaluating the system's effectiveness in fault detection through quantitative metrics; and 4) comparing the new system’s performance against conventional seismic imaging techniques. The research adopts a quantitative, experimental design within a geophysical field setting, utilizing seismic data collected from a 50-square-kilometer geological survey area known for complex faulting; the dataset comprises 150 contiguous seismic lines, with each line sampled at 4 ms intervals, totaling approximately 900,000 data points. A stratified random sampling method ensures representative data subsets for algorithm testing. The core data collection instruments include high-resolution 3D seismic sources and receiver arrays, along with GPS timing systems for precise spatial referencing. Data preprocessing involves noise attenuation through adaptive filtering, multiple attenuation, and amplitude corrections. The development of the imaging system employs advanced signal processing techniques such as full-waveform inversion (FWI), coherency calculations, and spectral decomposition, utilizing MATLAB and Python-based custom algorithms. The system’s detection capabilities are validated through controlled synthetic datasets with known fault parameters, and real datasets are analyzed via statistical measures including seismic attribute analysis and receiver function analysis. The analytical approach encompasses multiple methods quantitative assessment using receiver operating characteristic (ROC) curves to evaluate fault detection sensitivity and specificity, and statistical significance testing through paired t-tests comparing the performance of the new system with traditional migrated images. Additionally, a multivariate regression analysis explores the influence of geological complexity, data noise, and processing parameters on detection accuracy. Effectiveness is further examined through 3D visualization comparisons, and fault geometry parameters are extracted using automated fault tracking algorithms. Expected findings include a demonstrable improvement in fault delineation resolution, increased detection sensitivity, and reduced false-positive rates relative to traditional seismic imaging methods, supported by quantitative performance metrics and visualization clarity. The study significantly contributes to current geophysical knowledge by offering an integrated, systematically validated 3D seismic imaging framework capable of more precise fault detection, thereby enhancing hazard assessment and resource exploration strategies. It bridges the gap between conventional seismic imaging limitations and the need for high-resolution fault delineation, employing theoretical insights from wave physics and image processing theories, notably the wave equation migration and inverse theory. The research further substantiates the applicability of machine learning-enhanced imaging in structural fault detection, aligning with contemporary developments in geophysical data science. In conclusion, the developed 3D seismic imaging system demonstrates a robust enhancement in fault detection accuracy and resolution, serving as an effective tool for seismic hazard analysis and subsurface characterization. Recommendations include further algorithm refinement to accommodate varying geological conditions, integration with real-time data acquisition, and adaptation in offshore seismic surveys. Future studies are suggested to explore machine learning methodologies such as deep convolutional neural networks for automated fault recognition and to validate the system across diverse geological terrains to establish broader applicability and scalability.