Stochastic fault modelling using graph formalism for mining data.

@inproceedings{Fratani2026RM,
 abstract = {Determining whether distinct fault observations belong to the same fault can be addressed using a potential graph, in which nodes represent the fault observations and edges carry the potential that two observations belong to the same fault. This graph is computed based on expert rules and can be sampled stochastically to generate different fault correlations. Whereas this formalism has been demonstrated on fault traces identified on 2D sections, its application to other types of data has remained cursory. Here, this graph formalism is completed to handle small-scale drillholes data and large-scale reflection seismic data. For this, several new expert rules are proposed to consider information either from borehole images (position, orientation and thickness of the fault) or from 3D seismic cubes (fault likelihood). Stochastic construction of fault correlation is proposed with a new algorithm that can handle large number of observations such as typically available in mining or geotechnical surveys. This paper is an upgrade of last year’s contribution (Fratani et al., 2025). It includes updated fault correlation rules, new rules to integrate volumetric fault likelihood data.},
 author = {Fratani, Amandine and Caumon, Guillaume and Antoine, Christophe and Stoica, Radu and Baville, Romain and Prest, Chiara-Luna and Cabane, Antoine and Benguettaf, Bilal},
 booktitle = {2026 {RING} meeting},
 language = {en},
 pages = {57--71},
 publisher = {ASGA},
 title = {Stochastic fault modelling using graph formalism for mining data.},
 year = {2026}
}