AI-based Fault Detection in Synthetic Seismic Data for Hydrocarbon Exploration

Automatic fault detection in seismic data is pivotal for advancing hydrocarbon exploration and geological risk assessment. This study introduces an artificial intelligence (AI)-based framework leveraging the Random Forest (RF) algorithm to achieve accurate fault detection in synthetic seismic data. A dataset of ۲۰۰۰ samples, featuring coherence, dip angle, and curvature, was generated using a normal distribution and preprocessed through standardization, Z-score-based outlier removal, and SMOTE class balancing. Optimized via GridSearchCV, the RF model attained an accuracy of ۰.۶۷, an area under the receiver operating characteristic (ROC) curve (AUC) of ۰.۷۴, and a recall of ۰.۷۳ for the fault class, demonstrating robust detection capability. Visual analyses confirmed the effective separability of the selected features. This framework outperforms traditional manual interpretation, offering transformative applications in hydrocarbon exploration, geological structure analysis, and seismic risk assessment. Future work should validate this approach with real seismic data and incorporate advanced geophysical attributes to enhance generalizability. Overall, this study highlights a pivotal step toward automating fault detection, substantially improving the efficiency and accuracy of geophysical exploration.

کلیدواژه ها:

Fault detection ، artificial intelligence ، Random Forest ، synthetic seismic data ، hydrocarbon exploration