On the use of graph geometrical embedding and multi-dimensional analysis to compare the structure of well correlations and fault observation associations

@inproceedings{Baville2025RM,
 abstract = {The association between geological observations is a key step within the general subsurface modeling workflow. Two examples of associations are (1) fault observation associations and (2) well correlations. Graph is a mathematical object, whose architecture may be used to represent those associations. The aim of this work is to propose a methodology to compare multiple graphs (e.g., fault observation association graphs, and well correlation graphs), and to generate clusters of association graphs. A fault observation association generated by RINGral (a python software developed by RING Team to associate fault observations from seismic lines using random forest algorithm) can be represented as a graph whose nodes are the observations and edges represent their spatial connection. A well-marker correlation generated by WeCo (a C++/python software developed by RING Team to simulate multi-well correlation using dynamic time warping algorithm) can be represented as a directed acyclic graph whose nodes correspond to the association between well markers and edges represent the transitions between well marker associations. In this work, we use the Hausdorff distance to compute pairwise distances between graphs to identify graph clusters of (1) fault observation associations or (2) well-marker correlations based on their structures.},
 author = {Baville, Paul and Fratani, Amandine and Madelaine, Colin},
 booktitle = {2025 {RING} meeting},
 language = {en},
 pages = {197--208},
 publisher = {ASGA},
 title = {On the use of graph geometrical embedding and multi-dimensional analysis to compare the structure of well correlations and fault observation associations},
 year = {2025}
}