Comparative Study of Geologically Constrained and Random Sample Generation Methods

The selection of fault sample generation methods directly impacts the reliability of intelligent structural prediction models. To address the lack of geological regularity constraints in traditional random sample generation methods, this study systematically compares the differences between geologically constrained and random fault sample generation approaches, focusing on extensional structures within a deltaic sedimentary setting. By testing prediction models, we analyze the generated fault samples in terms of structural development patterns and geometric characteristics. The results show that: (1) Samples generated using geological constraints effectively conform to extensional structural features in deltaic environments, preserving fault segmentation and dip variation consistent with structural evolution patterns; (2) Random generation methods tend to produce geometrically unrealistic fault assemblages in complex structural settings. This research provides critical theoretical support for selecting fault sample generation methods in intelligent structural prediction and offers important guidance for improving interpretation accuracy.