Fault plane geometry reconstruction and fault region modeling method based on spatial topological constraints

CN122244360APending Publication Date: 2026-06-19CHINA UNIV OF GEOSCIENCES (WUHAN)

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: CHINA UNIV OF GEOSCIENCES (WUHAN)
Filing Date: 2026-04-15
Publication Date: 2026-06-19

AI Technical Summary

Technical Problem

Existing implicit modeling and GNN-based 3D geological modeling suffer from problems in fault structure modeling, such as difficulty in extracting fault plane topology, cross-fault feature mixing, difficulty in balancing high accuracy with global topological consistency, and excessive smoothing effects.

Method used

An implicit function based on radial basis functions is used to construct a geometric model of the fault plane. The fault mask adjacency matrix is combined to cut off the connection between nodes across the fault and construct a disconnected subgraph. A semi-supervised classification model is used for message passing and node classification. An inverse hierarchical linearization activation control strategy is adopted to achieve both high accuracy of the fault region and global topological consistency.

Benefits of technology

It achieves high-precision reconstruction of fault planes, eliminates cross-fault feature contamination, maintains the overall coherence of the model, alleviates the over-smoothing effect, and improves the automation and accuracy of 3D geological modeling.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122244360A_ABST

Patent Text Reader

Abstract

This application provides a method for fault plane geometry reconstruction and fault region modeling based on spatial topological constraints, relating to the field of 3D geological modeling. The method includes: acquiring a set of fault line points from geological data; after smoothing and normal vector extraction; constructing an implicit function that integrates position and direction constraints based on radial basis functions; solving for the zero level set to obtain the fault plane geometry model; calculating node symbol distances based on this model; constructing a fault mask adjacency matrix to cut off cross-fault edge connections, forming a disconnected subgraph; constructing a multi-layer graph neural network semi-supervised classification model, using linear transformations for shallow layers and nonlinear activation for the last layer to achieve piecewise linear expression; inputting the disconnected subgraph into the model for message passing and node classification, outputting a 3D geological model of the fault region. This application effectively solves the problems of cross-fault feature mixing, boundary ambiguity, and excessive smoothing through spatial topological constraints, achieving high-precision, globally consistent fault region modeling.

Need to check novelty before this filing date? Find Prior Art